summary refs log tree commit diff
diff options
context:
space:
mode:
-rw-r--r--Dockerfile22
-rw-r--r--changelog.d/3384.misc1
-rw-r--r--changelog.d/3543.misc1
-rw-r--r--changelog.d/3612.misc1
-rw-r--r--changelog.d/3626.bugfix1
-rw-r--r--changelog.d/3628.misc1
-rw-r--r--changelog.d/3634.misc1
-rw-r--r--contrib/docker/README.md8
-rw-r--r--contrib/docker/docker-compose.yml1
-rw-r--r--contrib/grafana/README.md6
-rw-r--r--contrib/grafana/synapse.json4961
-rw-r--r--synapse/federation/federation_server.py4
-rw-r--r--synapse/federation/send_queue.py63
-rw-r--r--synapse/federation/transaction_queue.py32
-rw-r--r--synapse/federation/transport/server.py3
-rw-r--r--synapse/federation/units.py1
-rw-r--r--synapse/handlers/federation.py28
-rw-r--r--synapse/secrets.py7
-rw-r--r--synapse/storage/__init__.py2
-rw-r--r--synapse/storage/appservice.py2
-rw-r--r--synapse/storage/event_federation.py2
-rw-r--r--synapse/storage/events.py6
-rw-r--r--synapse/storage/roommember.py2
-rw-r--r--synapse/storage/stream.py2
-rw-r--r--synapse/util/caches/descriptors.py131
-rw-r--r--tests/handlers/test_typing.py1
-rw-r--r--tests/util/caches/test_descriptors.py101
27 files changed, 5193 insertions, 198 deletions
diff --git a/Dockerfile b/Dockerfile
index 565341fee3..0242be5f68 100644
--- a/Dockerfile
+++ b/Dockerfile
@@ -1,16 +1,32 @@
 FROM docker.io/python:2-alpine3.7
 
-RUN apk add --no-cache --virtual .nacl_deps su-exec build-base libffi-dev zlib-dev libressl-dev libjpeg-turbo-dev linux-headers postgresql-dev libxslt-dev
+RUN apk add --no-cache --virtual .nacl_deps \
+        build-base \
+        libffi-dev \
+        libjpeg-turbo-dev \
+        libressl-dev \
+        libxslt-dev \
+        linux-headers \
+        postgresql-dev \
+        su-exec \
+        zlib-dev
 
 COPY . /synapse
 
 # A wheel cache may be provided in ./cache for faster build
 RUN cd /synapse \
- && pip install --upgrade pip setuptools psycopg2 lxml \
+ && pip install --upgrade \
+        lxml \
+        pip \
+        psycopg2 \
+        setuptools \
  && mkdir -p /synapse/cache \
  && pip install -f /synapse/cache --upgrade --process-dependency-links . \
  && mv /synapse/contrib/docker/start.py /synapse/contrib/docker/conf / \
- && rm -rf setup.py setup.cfg synapse
+ && rm -rf \
+        setup.cfg \
+        setup.py \
+        synapse
 
 VOLUME ["/data"]
 
diff --git a/changelog.d/3384.misc b/changelog.d/3384.misc
new file mode 100644
index 0000000000..5d56c876d9
--- /dev/null
+++ b/changelog.d/3384.misc
@@ -0,0 +1 @@
+Rewrite cache list decorator
diff --git a/changelog.d/3543.misc b/changelog.d/3543.misc
new file mode 100644
index 0000000000..d231d17749
--- /dev/null
+++ b/changelog.d/3543.misc
@@ -0,0 +1 @@
+Improve Dockerfile and docker-compose instructions
diff --git a/changelog.d/3612.misc b/changelog.d/3612.misc
new file mode 100644
index 0000000000..f90d2f2ff5
--- /dev/null
+++ b/changelog.d/3612.misc
@@ -0,0 +1 @@
+Make EventStore inherit from EventFederationStore
diff --git a/changelog.d/3626.bugfix b/changelog.d/3626.bugfix
new file mode 100644
index 0000000000..9a4b878986
--- /dev/null
+++ b/changelog.d/3626.bugfix
@@ -0,0 +1 @@
+Only import secrets when available (fix for py < 3.6)
diff --git a/changelog.d/3628.misc b/changelog.d/3628.misc
new file mode 100644
index 0000000000..1aebefbe18
--- /dev/null
+++ b/changelog.d/3628.misc
@@ -0,0 +1 @@
+Remove unused field "pdu_failures" from transactions.
diff --git a/changelog.d/3634.misc b/changelog.d/3634.misc
new file mode 100644
index 0000000000..2cd6af91ff
--- /dev/null
+++ b/changelog.d/3634.misc
@@ -0,0 +1 @@
+rename replication_layer to federation_client
diff --git a/contrib/docker/README.md b/contrib/docker/README.md
index 61592109cb..562cdaac2b 100644
--- a/contrib/docker/README.md
+++ b/contrib/docker/README.md
@@ -9,13 +9,7 @@ use that server.
 
 ## Build
 
-Build the docker image with the `docker build` command from the root of the synapse repository.
-
-```
-docker build -t docker.io/matrixdotorg/synapse .
-```
-
-The `-t` option sets the image tag. Official images are tagged `matrixdotorg/synapse:<version>` where `<version>` is the same as the release tag in the synapse git repository.
+Build the docker image with the `docker-compose build` command.
 
 You may have a local Python wheel cache available, in which case copy the relevant packages in the ``cache/`` directory at the root of the project.
 
diff --git a/contrib/docker/docker-compose.yml b/contrib/docker/docker-compose.yml
index 0b531949e0..3a8dfbae34 100644
--- a/contrib/docker/docker-compose.yml
+++ b/contrib/docker/docker-compose.yml
@@ -6,6 +6,7 @@ version: '3'
 services:
 
   synapse:
+    build: ../..
     image: docker.io/matrixdotorg/synapse:latest
     # Since snyapse does not retry to connect to the database, restart upon
     # failure
diff --git a/contrib/grafana/README.md b/contrib/grafana/README.md
new file mode 100644
index 0000000000..6a6cc0bed4
--- /dev/null
+++ b/contrib/grafana/README.md
@@ -0,0 +1,6 @@
+# Using the Synapse Grafana dashboard
+
+0. Set up Prometheus and Grafana. Out of scope for this readme. Useful documentation about using Grafana with Prometheus: http://docs.grafana.org/features/datasources/prometheus/
+1. Have your Prometheus scrape your Synapse. https://github.com/matrix-org/synapse/blob/master/docs/metrics-howto.rst
+2. Import dashboard into Grafana. Download `synapse.json`. Import it to Grafana and select the correct Prometheus datasource. http://docs.grafana.org/reference/export_import/
+3. Set up additional recording rules
diff --git a/contrib/grafana/synapse.json b/contrib/grafana/synapse.json
new file mode 100644
index 0000000000..94a1de58f4
--- /dev/null
+++ b/contrib/grafana/synapse.json
@@ -0,0 +1,4961 @@
+{
+  "__inputs": [
+    {
+      "name": "DS_PROMETHEUS",
+      "label": "Prometheus",
+      "description": "",
+      "type": "datasource",
+      "pluginId": "prometheus",
+      "pluginName": "Prometheus"
+    }
+  ],
+  "__requires": [
+    {
+      "type": "grafana",
+      "id": "grafana",
+      "name": "Grafana",
+      "version": "5.2.0"
+    },
+    {
+      "type": "panel",
+      "id": "graph",
+      "name": "Graph",
+      "version": "5.0.0"
+    },
+    {
+      "type": "panel",
+      "id": "heatmap",
+      "name": "Heatmap",
+      "version": "5.0.0"
+    },
+    {
+      "type": "datasource",
+      "id": "prometheus",
+      "name": "Prometheus",
+      "version": "5.0.0"
+    }
+  ],
+  "annotations": {
+    "list": [
+      {
+        "builtIn": 1,
+        "datasource": "$datasource",
+        "enable": false,
+        "hide": true,
+        "iconColor": "rgba(0, 211, 255, 1)",
+        "limit": 100,
+        "name": "Annotations & Alerts",
+        "showIn": 0,
+        "type": "dashboard"
+      }
+    ]
+  },
+  "editable": true,
+  "gnetId": null,
+  "graphTooltip": 0,
+  "id": null,
+  "iteration": 1533026624326,
+  "links": [
+    {
+      "asDropdown": true,
+      "icon": "external link",
+      "keepTime": true,
+      "tags": [
+        "matrix"
+      ],
+      "title": "Dashboards",
+      "type": "dashboards"
+    }
+  ],
+  "panels": [
+    {
+      "collapsed": false,
+      "gridPos": {
+        "h": 1,
+        "w": 24,
+        "x": 0,
+        "y": 0
+      },
+      "id": 73,
+      "panels": [],
+      "title": "Overview",
+      "type": "row"
+    },
+    {
+      "aliasColors": {},
+      "bars": false,
+      "dashLength": 10,
+      "dashes": false,
+      "datasource": "${DS_PROMETHEUS}",
+      "fill": 1,
+      "gridPos": {
+        "h": 9,
+        "w": 12,
+        "x": 0,
+        "y": 1
+      },
+      "id": 75,
+      "legend": {
+        "avg": false,
+        "current": false,
+        "max": false,
+        "min": false,
+        "show": true,
+        "total": false,
+        "values": false
+      },
+      "lines": true,
+      "linewidth": 1,
+      "links": [],
+      "nullPointMode": "null",
+      "percentage": false,
+      "pointradius": 5,
+      "points": false,
+      "renderer": "flot",
+      "seriesOverrides": [],
+      "spaceLength": 10,
+      "stack": false,
+      "steppedLine": false,
+      "targets": [
+        {
+          "expr": "process_cpu_seconds:rate2m{instance=\"$instance\",job=~\"$job\",index=~\"$index\"}",
+          "format": "time_series",
+          "intervalFactor": 1,
+          "legendFormat": "{{job}}-{{index}} ",
+          "refId": "A"
+        }
+      ],
+      "thresholds": [],
+      "timeFrom": null,
+      "timeShift": null,
+      "title": "CPU usage",
+      "tooltip": {
+        "shared": true,
+        "sort": 0,
+        "value_type": "individual"
+      },
+      "type": "graph",
+      "xaxis": {
+        "buckets": null,
+        "mode": "time",
+        "name": null,
+        "show": true,
+        "values": []
+      },
+      "yaxes": [
+        {
+          "decimals": null,
+          "format": "percentunit",
+          "label": null,
+          "logBase": 1,
+          "max": "1",
+          "min": "0",
+          "show": true
+        },
+        {
+          "format": "short",
+          "label": null,
+          "logBase": 1,
+          "max": null,
+          "min": null,
+          "show": true
+        }
+      ],
+      "yaxis": {
+        "align": false,
+        "alignLevel": null
+      }
+    },
+    {
+      "cards": {
+        "cardPadding": 0,
+        "cardRound": null
+      },
+      "color": {
+        "cardColor": "#b4ff00",
+        "colorScale": "sqrt",
+        "colorScheme": "interpolateSpectral",
+        "exponent": 0.5,
+        "mode": "spectrum"
+      },
+      "dataFormat": "tsbuckets",
+      "datasource": "${DS_PROMETHEUS}",
+      "gridPos": {
+        "h": 9,
+        "w": 12,
+        "x": 12,
+        "y": 1
+      },
+      "heatmap": {},
+      "highlightCards": true,
+      "id": 85,
+      "legend": {
+        "show": false
+      },
+      "links": [],
+      "targets": [
+        {
+          "expr": "sum(rate(synapse_http_server_response_time_seconds_bucket{servlet='RoomSendEventRestServlet',instance=\"$instance\"}[$bucket_size])) by (le)",
+          "format": "heatmap",
+          "intervalFactor": 1,
+          "legendFormat": "{{le}}",
+          "refId": "A"
+        }
+      ],
+      "title": "Event Send Time",
+      "tooltip": {
+        "show": true,
+        "showHistogram": false
+      },
+      "type": "heatmap",
+      "xAxis": {
+        "show": true
+      },
+      "xBucketNumber": null,
+      "xBucketSize": null,
+      "yAxis": {
+        "decimals": null,
+        "format": "s",
+        "logBase": 2,
+        "max": null,
+        "min": null,
+        "show": true,
+        "splitFactor": null
+      },
+      "yBucketBound": "auto",
+      "yBucketNumber": null,
+      "yBucketSize": null
+    },
+    {
+      "aliasColors": {},
+      "bars": false,
+      "dashLength": 10,
+      "dashes": false,
+      "datasource": "$datasource",
+      "editable": true,
+      "error": false,
+      "fill": 1,
+      "grid": {},
+      "gridPos": {
+        "h": 7,
+        "w": 12,
+        "x": 0,
+        "y": 10
+      },
+      "id": 33,
+      "legend": {
+        "avg": false,
+        "current": false,
+        "max": false,
+        "min": false,
+        "show": false,
+        "total": false,
+        "values": false
+      },
+      "lines": true,
+      "linewidth": 2,
+      "links": [],
+      "nullPointMode": "null",
+      "percentage": false,
+      "pointradius": 5,
+      "points": false,
+      "renderer": "flot",
+      "seriesOverrides": [],
+      "spaceLength": 10,
+      "stack": false,
+      "steppedLine": false,
+      "targets": [
+        {
+          "expr": "sum(rate(synapse_storage_events_persisted_events{instance=\"$instance\"}[$bucket_size])) without (job,index)",
+          "format": "time_series",
+          "intervalFactor": 2,
+          "legendFormat": "",
+          "refId": "A",
+          "step": 20,
+          "target": ""
+        }
+      ],
+      "thresholds": [],
+      "timeFrom": null,
+      "timeShift": null,
+      "title": "Events Persisted",
+      "tooltip": {
+        "shared": true,
+        "sort": 0,
+        "value_type": "cumulative"
+      },
+      "type": "graph",
+      "xaxis": {
+        "buckets": null,
+        "mode": "time",
+        "name": null,
+        "show": true,
+        "values": []
+      },
+      "yaxes": [
+        {
+          "format": "hertz",
+          "logBase": 1,
+          "max": null,
+          "min": null,
+          "show": true
+        },
+        {
+          "format": "short",
+          "logBase": 1,
+          "max": null,
+          "min": null,
+          "show": true
+        }
+      ],
+      "yaxis": {
+        "align": false,
+        "alignLevel": null
+      }
+    },
+    {
+      "collapsed": true,
+      "gridPos": {
+        "h": 1,
+        "w": 24,
+        "x": 0,
+        "y": 17
+      },
+      "id": 54,
+      "panels": [
+        {
+          "aliasColors": {},
+          "bars": false,
+          "dashLength": 10,
+          "dashes": false,
+          "datasource": "$datasource",
+          "editable": true,
+          "error": false,
+          "fill": 0,
+          "grid": {},
+          "gridPos": {
+            "h": 7,
+            "w": 12,
+            "x": 0,
+            "y": 18
+          },
+          "id": 34,
+          "legend": {
+            "avg": false,
+            "current": false,
+            "max": false,
+            "min": false,
+            "show": true,
+            "total": false,
+            "values": false
+          },
+          "lines": true,
+          "linewidth": 2,
+          "links": [],
+          "nullPointMode": "null",
+          "percentage": false,
+          "pointradius": 5,
+          "points": false,
+          "renderer": "flot",
+          "seriesOverrides": [],
+          "spaceLength": 10,
+          "stack": false,
+          "steppedLine": true,
+          "targets": [
+            {
+              "expr": "process_resident_memory_bytes{instance=\"$instance\",job=~\"$job\",index=~\"$index\"}",
+              "format": "time_series",
+              "intervalFactor": 2,
+              "legendFormat": "{{job}} {{index}}",
+              "refId": "A",
+              "step": 20,
+              "target": ""
+            }
+          ],
+          "thresholds": [],
+          "timeFrom": null,
+          "timeShift": null,
+          "title": "Memory",
+          "tooltip": {
+            "shared": true,
+            "sort": 0,
+            "value_type": "cumulative"
+          },
+          "type": "graph",
+          "xaxis": {
+            "buckets": null,
+            "mode": "time",
+            "name": null,
+            "show": true,
+            "values": []
+          },
+          "yaxes": [
+            {
+              "format": "bytes",
+              "logBase": 1,
+              "max": null,
+              "min": "0",
+              "show": true
+            },
+            {
+              "format": "short",
+              "logBase": 1,
+              "max": null,
+              "min": null,
+              "show": true
+            }
+          ],
+          "yaxis": {
+            "align": false,
+            "alignLevel": null
+          }
+        },
+        {
+          "aliasColors": {},
+          "bars": false,
+          "dashLength": 10,
+          "dashes": false,
+          "datasource": "$datasource",
+          "fill": 1,
+          "gridPos": {
+            "h": 7,
+            "w": 12,
+            "x": 12,
+            "y": 18
+          },
+          "id": 37,
+          "legend": {
+            "avg": false,
+            "current": false,
+            "max": false,
+            "min": false,
+            "show": true,
+            "total": false,
+            "values": false
+          },
+          "lines": true,
+          "linewidth": 1,
+          "links": [],
+          "nullPointMode": "null",
+          "percentage": false,
+          "pointradius": 5,
+          "points": false,
+          "renderer": "flot",
+          "seriesOverrides": [
+            {
+              "alias": "/max$/",
+              "color": "#890F02",
+              "fill": 0,
+              "legend": false
+            }
+          ],
+          "spaceLength": 10,
+          "stack": false,
+          "steppedLine": false,
+          "targets": [
+            {
+              "expr": "process_open_fds{instance=\"$instance\",job=~\"$job\",index=~\"$index\"}",
+              "format": "time_series",
+              "hide": false,
+              "intervalFactor": 2,
+              "legendFormat": "{{job}}-{{index}}",
+              "refId": "A",
+              "step": 20
+            },
+            {
+              "expr": "process_max_fds{instance=\"$instance\",job=~\"$job\",index=~\"$index\"}",
+              "format": "time_series",
+              "hide": true,
+              "intervalFactor": 2,
+              "legendFormat": "{{job}}-{{index}} max",
+              "refId": "B",
+              "step": 20
+            }
+          ],
+          "thresholds": [],
+          "timeFrom": null,
+          "timeShift": null,
+          "title": "Open FDs",
+          "tooltip": {
+            "shared": true,
+            "sort": 0,
+            "value_type": "individual"
+          },
+          "type": "graph",
+          "xaxis": {
+            "buckets": null,
+            "mode": "time",
+            "name": null,
+            "show": true,
+            "values": []
+          },
+          "yaxes": [
+            {
+              "format": "none",
+              "label": null,
+              "logBase": 1,
+              "max": null,
+              "min": null,
+              "show": true
+            },
+            {
+              "format": "short",
+              "label": null,
+              "logBase": 1,
+              "max": null,
+              "min": null,
+              "show": true
+            }
+          ],
+          "yaxis": {
+            "align": false,
+            "alignLevel": null
+          }
+        },
+        {
+          "aliasColors": {},
+          "bars": false,
+          "dashLength": 10,
+          "dashes": false,
+          "datasource": "$datasource",
+          "fill": 1,
+          "gridPos": {
+            "h": 7,
+            "w": 12,
+            "x": 0,
+            "y": 25
+          },
+          "id": 48,
+          "legend": {
+            "avg": false,
+            "current": false,
+            "max": false,
+            "min": false,
+            "show": true,
+            "total": false,
+            "values": false
+          },
+          "lines": true,
+          "linewidth": 1,
+          "links": [],
+          "nullPointMode": "null",
+          "percentage": false,
+          "pointradius": 5,
+          "points": false,
+          "renderer": "flot",
+          "seriesOverrides": [],
+          "spaceLength": 10,
+          "stack": false,
+          "steppedLine": false,
+          "targets": [
+            {
+              "expr": "rate(synapse_storage_schedule_time_sum{instance=\"$instance\",job=~\"$job\",index=~\"$index\"}[$bucket_size])/rate(synapse_storage_schedule_time_count[$bucket_size])",
+              "format": "time_series",
+              "intervalFactor": 2,
+              "legendFormat": "{{job}}-{{index}}",
+              "refId": "A",
+              "step": 20
+            }
+          ],
+          "thresholds": [],
+          "timeFrom": null,
+          "timeShift": null,
+          "title": "Avg time waiting for db conn",
+          "tooltip": {
+            "shared": true,
+            "sort": 0,
+            "value_type": "individual"
+          },
+          "type": "graph",
+          "xaxis": {
+            "buckets": null,
+            "mode": "time",
+            "name": null,
+            "show": true,
+            "values": []
+          },
+          "yaxes": [
+            {
+              "decimals": null,
+              "format": "s",
+              "label": "",
+              "logBase": 1,
+              "max": null,
+              "min": "0",
+              "show": true
+            },
+            {
+              "format": "short",
+              "label": null,
+              "logBase": 1,
+              "max": null,
+              "min": null,
+              "show": false
+            }
+          ],
+          "yaxis": {
+            "align": false,
+            "alignLevel": null
+          }
+        },
+        {
+          "aliasColors": {},
+          "bars": false,
+          "dashLength": 10,
+          "dashes": false,
+          "datasource": "$datasource",
+          "fill": 1,
+          "gridPos": {
+            "h": 7,
+            "w": 12,
+            "x": 12,
+            "y": 25
+          },
+          "id": 49,
+          "legend": {
+            "avg": false,
+            "current": false,
+            "max": false,
+            "min": false,
+            "show": true,
+            "total": false,
+            "values": false
+          },
+          "lines": true,
+          "linewidth": 1,
+          "links": [],
+          "nullPointMode": "null",
+          "percentage": false,
+          "pointradius": 5,
+          "points": false,
+          "renderer": "flot",
+          "seriesOverrides": [
+            {
+              "alias": "/^up/",
+              "legend": false,
+              "yaxis": 2
+            }
+          ],
+          "spaceLength": 10,
+          "stack": false,
+          "steppedLine": false,
+          "targets": [
+            {
+              "expr": "scrape_duration_seconds{instance=\"$instance\",job=~\"$job\",index=~\"$index\"}",
+              "format": "time_series",
+              "interval": "",
+              "intervalFactor": 2,
+              "legendFormat": "{{job}}-{{index}}",
+              "refId": "A",
+              "step": 20
+            }
+          ],
+          "thresholds": [],
+          "timeFrom": null,
+          "timeShift": null,
+          "title": "Prometheus scrape time",
+          "tooltip": {
+            "shared": true,
+            "sort": 0,
+            "value_type": "individual"
+          },
+          "type": "graph",
+          "xaxis": {
+            "buckets": null,
+            "mode": "time",
+            "name": null,
+            "show": true,
+            "values": []
+          },
+          "yaxes": [
+            {
+              "format": "s",
+              "label": null,
+              "logBase": 1,
+              "max": null,
+              "min": "0",
+              "show": true
+            },
+            {
+              "decimals": 0,
+              "format": "none",
+              "label": "",
+              "logBase": 1,
+              "max": "0",
+              "min": "-1",
+              "show": false
+            }
+          ],
+          "yaxis": {
+            "align": false,
+            "alignLevel": null
+          }
+        },
+        {
+          "aliasColors": {},
+          "bars": false,
+          "dashLength": 10,
+          "dashes": false,
+          "datasource": "$datasource",
+          "fill": 1,
+          "gridPos": {
+            "h": 7,
+            "w": 12,
+            "x": 0,
+            "y": 32
+          },
+          "id": 50,
+          "legend": {
+            "avg": false,
+            "current": false,
+            "max": false,
+            "min": false,
+            "show": true,
+            "total": false,
+            "values": false
+          },
+          "lines": true,
+          "linewidth": 1,
+          "links": [],
+          "nullPointMode": "null",
+          "percentage": false,
+          "pointradius": 5,
+          "points": false,
+          "renderer": "flot",
+          "seriesOverrides": [],
+          "spaceLength": 10,
+          "stack": false,
+          "steppedLine": false,
+          "targets": [
+            {
+              "expr": "rate(python_twisted_reactor_tick_time_sum{instance=\"$instance\",job=~\"$job\",index=~\"$index\"}[$bucket_size])/rate(python_twisted_reactor_tick_time_count[$bucket_size])",
+              "format": "time_series",
+              "interval": "",
+              "intervalFactor": 2,
+              "legendFormat": "{{job}}-{{index}}",
+              "refId": "A",
+              "step": 20
+            }
+          ],
+          "thresholds": [],
+          "timeFrom": null,
+          "timeShift": null,
+          "title": "Avg reactor tick time",
+          "tooltip": {
+            "shared": true,
+            "sort": 0,
+            "value_type": "individual"
+          },
+          "type": "graph",
+          "xaxis": {
+            "buckets": null,
+            "mode": "time",
+            "name": null,
+            "show": true,
+            "values": []
+          },
+          "yaxes": [
+            {
+              "format": "s",
+              "label": null,
+              "logBase": 1,
+              "max": null,
+              "min": null,
+              "show": true
+            },
+            {
+              "format": "short",
+              "label": null,
+              "logBase": 1,
+              "max": null,
+              "min": null,
+              "show": false
+            }
+          ],
+          "yaxis": {
+            "align": false,
+            "alignLevel": null
+          }
+        },
+        {
+          "aliasColors": {},
+          "bars": false,
+          "dashLength": 10,
+          "dashes": false,
+          "datasource": "$datasource",
+          "editable": true,
+          "error": false,
+          "fill": 1,
+          "grid": {},
+          "gridPos": {
+            "h": 7,
+            "w": 12,
+            "x": 12,
+            "y": 32
+          },
+          "id": 5,
+          "legend": {
+            "alignAsTable": false,
+            "avg": false,
+            "current": false,
+            "hideEmpty": false,
+            "hideZero": false,
+            "max": false,
+            "min": false,
+            "rightSide": false,
+            "show": true,
+            "total": false,
+            "values": false
+          },
+          "lines": true,
+          "linewidth": 1,
+          "links": [],
+          "nullPointMode": "null",
+          "percentage": false,
+          "pointradius": 5,
+          "points": false,
+          "renderer": "flot",
+          "seriesOverrides": [
+            {
+              "alias": "/user/"
+            },
+            {
+              "alias": "/system/"
+            }
+          ],
+          "spaceLength": 10,
+          "stack": false,
+          "steppedLine": false,
+          "targets": [
+            {
+              "expr": "rate(process_cpu_system_seconds_total{instance=\"$instance\",job=~\"$job\",index=~\"$index\"}[$bucket_size])",
+              "format": "time_series",
+              "intervalFactor": 1,
+              "legendFormat": "{{job}}-{{index}} system ",
+              "metric": "",
+              "refId": "B",
+              "step": 20
+            },
+            {
+              "expr": "rate(process_cpu_user_seconds_total{instance=\"$instance\",job=~\"$job\",index=~\"$index\"}[$bucket_size])",
+              "format": "time_series",
+              "hide": false,
+              "interval": "",
+              "intervalFactor": 1,
+              "legendFormat": "{{job}}-{{index}} user",
+              "refId": "A",
+              "step": 20
+            }
+          ],
+          "thresholds": [
+            {
+              "colorMode": "custom",
+              "line": true,
+              "lineColor": "rgba(216, 200, 27, 0.27)",
+              "op": "gt",
+              "value": 0.5
+            },
+            {
+              "colorMode": "custom",
+              "line": true,
+              "lineColor": "rgba(234, 112, 112, 0.22)",
+              "op": "gt",
+              "value": 0.8
+            }
+          ],
+          "timeFrom": null,
+          "timeShift": null,
+          "title": "CPU",
+          "tooltip": {
+            "shared": true,
+            "sort": 0,
+            "value_type": "individual"
+          },
+          "type": "graph",
+          "xaxis": {
+            "buckets": null,
+            "mode": "time",
+            "name": null,
+            "show": true,
+            "values": []
+          },
+          "yaxes": [
+            {
+              "decimals": null,
+              "format": "percentunit",
+              "label": "",
+              "logBase": 1,
+              "max": "1.2",
+              "min": 0,
+              "show": true
+            },
+            {
+              "format": "short",
+              "logBase": 1,
+              "max": null,
+              "min": null,
+              "show": true
+            }
+          ],
+          "yaxis": {
+            "align": false,
+            "alignLevel": null
+          }
+        },
+        {
+          "aliasColors": {},
+          "bars": false,
+          "dashLength": 10,
+          "dashes": false,
+          "datasource": "${DS_PROMETHEUS}",
+          "fill": 0,
+          "gridPos": {
+            "h": 7,
+            "w": 12,
+            "x": 0,
+            "y": 39
+          },
+          "id": 53,
+          "legend": {
+            "avg": false,
+            "current": false,
+            "max": false,
+            "min": false,
+            "show": true,
+            "total": false,
+            "values": false
+          },
+          "lines": true,
+          "linewidth": 1,
+          "links": [],
+          "nullPointMode": "null",
+          "percentage": false,
+          "pointradius": 5,
+          "points": false,
+          "renderer": "flot",
+          "seriesOverrides": [],
+          "spaceLength": 10,
+          "stack": false,
+          "steppedLine": false,
+          "targets": [
+            {
+              "expr": "min_over_time(up{instance=\"$instance\",job=~\"$job\",index=~\"$index\"}[$bucket_size])",
+              "format": "time_series",
+              "intervalFactor": 2,
+              "legendFormat": "{{job}}-{{index}}",
+              "refId": "A"
+            }
+          ],
+          "thresholds": [],
+          "timeFrom": null,
+          "timeShift": null,
+          "title": "Up",
+          "tooltip": {
+            "shared": true,
+            "sort": 0,
+            "value_type": "individual"
+          },
+          "type": "graph",
+          "xaxis": {
+            "buckets": null,
+            "mode": "time",
+            "name": null,
+            "show": true,
+            "values": []
+          },
+          "yaxes": [
+            {
+              "format": "short",
+              "label": null,
+              "logBase": 1,
+              "max": null,
+              "min": null,
+              "show": true
+            },
+            {
+              "format": "short",
+              "label": null,
+              "logBase": 1,
+              "max": null,
+              "min": null,
+              "show": true
+            }
+          ],
+          "yaxis": {
+            "align": false,
+            "alignLevel": null
+          }
+        }
+      ],
+      "repeat": null,
+      "title": "Process info",
+      "type": "row"
+    },
+    {
+      "collapsed": true,
+      "gridPos": {
+        "h": 1,
+        "w": 24,
+        "x": 0,
+        "y": 18
+      },
+      "id": 56,
+      "panels": [
+        {
+          "aliasColors": {},
+          "bars": false,
+          "dashLength": 10,
+          "dashes": false,
+          "datasource": "$datasource",
+          "decimals": 1,
+          "fill": 1,
+          "gridPos": {
+            "h": 7,
+            "w": 12,
+            "x": 0,
+            "y": 49
+          },
+          "id": 40,
+          "legend": {
+            "avg": false,
+            "current": false,
+            "max": false,
+            "min": false,
+            "show": true,
+            "total": false,
+            "values": false
+          },
+          "lines": true,
+          "linewidth": 1,
+          "links": [],
+          "nullPointMode": "null",
+          "percentage": false,
+          "pointradius": 5,
+          "points": false,
+          "renderer": "flot",
+          "seriesOverrides": [],
+          "spaceLength": 10,
+          "stack": false,
+          "steppedLine": false,
+          "targets": [
+            {
+              "expr": "rate(synapse_storage_events_persisted_by_source_type{instance=\"$instance\",job=~\"$job\",index=~\"$index\"}[$bucket_size])",
+              "format": "time_series",
+              "intervalFactor": 2,
+              "legendFormat": "{{type}}",
+              "refId": "D"
+            }
+          ],
+          "thresholds": [],
+          "timeFrom": null,
+          "timeShift": null,
+          "title": "Events/s Local vs Remote",
+          "tooltip": {
+            "shared": true,
+            "sort": 2,
+            "value_type": "individual"
+          },
+          "type": "graph",
+          "xaxis": {
+            "buckets": null,
+            "mode": "time",
+            "name": null,
+            "show": true,
+            "values": []
+          },
+          "yaxes": [
+            {
+              "format": "hertz",
+              "label": "",
+              "logBase": 1,
+              "max": null,
+              "min": "0",
+              "show": true
+            },
+            {
+              "format": "short",
+              "label": null,
+              "logBase": 1,
+              "max": null,
+              "min": null,
+              "show": true
+            }
+          ],
+          "yaxis": {
+            "align": false,
+            "alignLevel": null
+          }
+        },
+        {
+          "aliasColors": {},
+          "bars": false,
+          "dashLength": 10,
+          "dashes": false,
+          "datasource": "$datasource",
+          "decimals": 1,
+          "fill": 1,
+          "gridPos": {
+            "h": 7,
+            "w": 12,
+            "x": 12,
+            "y": 49
+          },
+          "id": 46,
+          "legend": {
+            "avg": false,
+            "current": false,
+            "max": false,
+            "min": false,
+            "show": true,
+            "total": false,
+            "values": false
+          },
+          "lines": true,
+          "linewidth": 1,
+          "links": [],
+          "nullPointMode": "null",
+          "percentage": false,
+          "pointradius": 5,
+          "points": false,
+          "renderer": "flot",
+          "seriesOverrides": [],
+          "spaceLength": 10,
+          "stack": false,
+          "steppedLine": false,
+          "targets": [
+            {
+              "expr": "rate(synapse_storage_events_persisted_by_event_type{job=~\"$job\",index=~\"$index\",instance=\"$instance\"}[$bucket_size])",
+              "format": "time_series",
+              "instant": false,
+              "intervalFactor": 2,
+              "legendFormat": "{{type}}",
+              "refId": "A",
+              "step": 20
+            }
+          ],
+          "thresholds": [],
+          "timeFrom": null,
+          "timeShift": null,
+          "title": "Events/s by Type",
+          "tooltip": {
+            "shared": false,
+            "sort": 2,
+            "value_type": "individual"
+          },
+          "type": "graph",
+          "xaxis": {
+            "buckets": null,
+            "mode": "time",
+            "name": null,
+            "show": true,
+            "values": []
+          },
+          "yaxes": [
+            {
+              "format": "hertz",
+              "label": null,
+              "logBase": 1,
+              "max": null,
+              "min": "0",
+              "show": true
+            },
+            {
+              "format": "short",
+              "label": null,
+              "logBase": 1,
+              "max": null,
+              "min": null,
+              "show": true
+            }
+          ],
+          "yaxis": {
+            "align": false,
+            "alignLevel": null
+          }
+        },
+        {
+          "aliasColors": {
+            "irc-freenode (local)": "#EAB839"
+          },
+          "bars": false,
+          "dashLength": 10,
+          "dashes": false,
+          "datasource": "$datasource",
+          "decimals": 1,
+          "fill": 1,
+          "gridPos": {
+            "h": 7,
+            "w": 12,
+            "x": 0,
+            "y": 56
+          },
+          "id": 44,
+          "legend": {
+            "alignAsTable": true,
+            "avg": false,
+            "current": false,
+            "hideEmpty": true,
+            "hideZero": true,
+            "max": false,
+            "min": false,
+            "show": true,
+            "total": false,
+            "values": false
+          },
+          "lines": true,
+          "linewidth": 1,
+          "links": [],
+          "nullPointMode": "null",
+          "percentage": false,
+          "pointradius": 5,
+          "points": false,
+          "renderer": "flot",
+          "seriesOverrides": [],
+          "spaceLength": 10,
+          "stack": false,
+          "steppedLine": false,
+          "targets": [
+            {
+              "expr": "rate(synapse_storage_events_persisted_by_origin{job=~\"$job\",index=~\"$index\",instance=\"$instance\"}[$bucket_size])",
+              "format": "time_series",
+              "intervalFactor": 2,
+              "legendFormat": "{{origin_entity}} ({{origin_type}})",
+              "refId": "A",
+              "step": 20
+            }
+          ],
+          "thresholds": [],
+          "timeFrom": null,
+          "timeShift": null,
+          "title": "Events/s by Origin",
+          "tooltip": {
+            "shared": false,
+            "sort": 2,
+            "value_type": "individual"
+          },
+          "type": "graph",
+          "xaxis": {
+            "buckets": null,
+            "mode": "time",
+            "name": null,
+            "show": true,
+            "values": []
+          },
+          "yaxes": [
+            {
+              "format": "hertz",
+              "label": null,
+              "logBase": 1,
+              "max": null,
+              "min": "0",
+              "show": true
+            },
+            {
+              "format": "short",
+              "label": null,
+              "logBase": 1,
+              "max": null,
+              "min": null,
+              "show": true
+            }
+          ],
+          "yaxis": {
+            "align": false,
+            "alignLevel": null
+          }
+        },
+        {
+          "aliasColors": {},
+          "bars": false,
+          "dashLength": 10,
+          "dashes": false,
+          "datasource": "$datasource",
+          "decimals": 1,
+          "fill": 1,
+          "gridPos": {
+            "h": 7,
+            "w": 12,
+            "x": 12,
+            "y": 56
+          },
+          "id": 45,
+          "legend": {
+            "alignAsTable": true,
+            "avg": false,
+            "current": false,
+            "hideEmpty": true,
+            "hideZero": true,
+            "max": false,
+            "min": false,
+            "show": true,
+            "total": false,
+            "values": false
+          },
+          "lines": true,
+          "linewidth": 1,
+          "links": [],
+          "nullPointMode": "null",
+          "percentage": false,
+          "pointradius": 5,
+          "points": false,
+          "renderer": "flot",
+          "seriesOverrides": [],
+          "spaceLength": 10,
+          "stack": false,
+          "steppedLine": false,
+          "targets": [
+            {
+              "expr": "sum(rate(synapse_storage_events_persisted_events_sep{job=~\"$job\",index=~\"$index\", type=\"m.room.member\",instance=\"$instance\"}[$bucket_size])) by (origin_type, origin_entity)",
+              "format": "time_series",
+              "intervalFactor": 2,
+              "legendFormat": "{{origin_entity}} ({{origin_type}})",
+              "refId": "A",
+              "step": 20
+            }
+          ],
+          "thresholds": [],
+          "timeFrom": null,
+          "timeShift": null,
+          "title": "Memberships/s by Origin",
+          "tooltip": {
+            "shared": true,
+            "sort": 2,
+            "value_type": "individual"
+          },
+          "type": "graph",
+          "xaxis": {
+            "buckets": null,
+            "mode": "time",
+            "name": null,
+            "show": true,
+            "values": []
+          },
+          "yaxes": [
+            {
+              "format": "hertz",
+              "label": null,
+              "logBase": 1,
+              "max": null,
+              "min": "0",
+              "show": true
+            },
+            {
+              "format": "short",
+              "label": null,
+              "logBase": 1,
+              "max": null,
+              "min": null,
+              "show": true
+            }
+          ],
+          "yaxis": {
+            "align": false,
+            "alignLevel": null
+          }
+        }
+      ],
+      "repeat": null,
+      "title": "Event persist rates",
+      "type": "row"
+    },
+    {
+      "collapsed": true,
+      "gridPos": {
+        "h": 1,
+        "w": 24,
+        "x": 0,
+        "y": 19
+      },
+      "id": 57,
+      "panels": [
+        {
+          "aliasColors": {},
+          "bars": false,
+          "dashLength": 10,
+          "dashes": false,
+          "datasource": "$datasource",
+          "decimals": null,
+          "editable": true,
+          "error": false,
+          "fill": 2,
+          "grid": {},
+          "gridPos": {
+            "h": 8,
+            "w": 12,
+            "x": 0,
+            "y": 48
+          },
+          "id": 4,
+          "legend": {
+            "alignAsTable": true,
+            "avg": false,
+            "current": false,
+            "hideEmpty": false,
+            "hideZero": true,
+            "max": false,
+            "min": false,
+            "rightSide": false,
+            "show": true,
+            "total": false,
+            "values": false
+          },
+          "lines": true,
+          "linewidth": 1,
+          "links": [],
+          "nullPointMode": "null",
+          "percentage": false,
+          "pointradius": 5,
+          "points": false,
+          "renderer": "flot",
+          "seriesOverrides": [],
+          "spaceLength": 10,
+          "stack": false,
+          "steppedLine": false,
+          "targets": [
+            {
+              "expr": "rate(synapse_http_server_requests_received{instance=\"$instance\",job=~\"$job\",index=~\"$index\"}[$bucket_size])",
+              "format": "time_series",
+              "interval": "",
+              "intervalFactor": 2,
+              "legendFormat": "{{job}}-{{index}} {{method}} {{servlet}} {{tag}}",
+              "refId": "A",
+              "step": 20
+            }
+          ],
+          "thresholds": [
+            {
+              "colorMode": "custom",
+              "fill": true,
+              "fillColor": "rgba(216, 200, 27, 0.27)",
+              "op": "gt",
+              "value": 100
+            },
+            {
+              "colorMode": "custom",
+              "fill": true,
+              "fillColor": "rgba(234, 112, 112, 0.22)",
+              "op": "gt",
+              "value": 250
+            }
+          ],
+          "timeFrom": null,
+          "timeShift": null,
+          "title": "Request Count by arrival time",
+          "tooltip": {
+            "shared": false,
+            "sort": 0,
+            "value_type": "individual"
+          },
+          "transparent": false,
+          "type": "graph",
+          "xaxis": {
+            "buckets": null,
+            "mode": "time",
+            "name": null,
+            "show": true,
+            "values": []
+          },
+          "yaxes": [
+            {
+              "format": "hertz",
+              "logBase": 1,
+              "max": null,
+              "min": null,
+              "show": true
+            },
+            {
+              "format": "short",
+              "logBase": 1,
+              "max": null,
+              "min": null,
+              "show": true
+            }
+          ],
+          "yaxis": {
+            "align": false,
+            "alignLevel": null
+          }
+        },
+        {
+          "aliasColors": {},
+          "bars": false,
+          "dashLength": 10,
+          "dashes": false,
+          "datasource": "$datasource",
+          "editable": true,
+          "error": false,
+          "fill": 1,
+          "grid": {},
+          "gridPos": {
+            "h": 8,
+            "w": 12,
+            "x": 12,
+            "y": 48
+          },
+          "id": 32,
+          "legend": {
+            "avg": false,
+            "current": false,
+            "max": false,
+            "min": false,
+            "show": true,
+            "total": false,
+            "values": false
+          },
+          "lines": true,
+          "linewidth": 2,
+          "links": [],
+          "nullPointMode": "null",
+          "percentage": false,
+          "pointradius": 5,
+          "points": false,
+          "renderer": "flot",
+          "seriesOverrides": [],
+          "spaceLength": 10,
+          "stack": false,
+          "steppedLine": false,
+          "targets": [
+            {
+              "expr": "rate(synapse_http_server_requests_received{instance=\"$instance\",job=~\"$job\",index=~\"$index\",method!=\"OPTIONS\"}[$bucket_size]) and topk(10,synapse_http_server_requests_received{instance=\"$instance\",job=~\"$job\",method!=\"OPTIONS\"})",
+              "format": "time_series",
+              "intervalFactor": 2,
+              "legendFormat": "{{method}} {{servlet}} {{job}}-{{index}}",
+              "refId": "A",
+              "step": 20,
+              "target": ""
+            }
+          ],
+          "thresholds": [],
+          "timeFrom": null,
+          "timeShift": null,
+          "title": "Top 10 Request Counts",
+          "tooltip": {
+            "shared": false,
+            "sort": 0,
+            "value_type": "cumulative"
+          },
+          "type": "graph",
+          "xaxis": {
+            "buckets": null,
+            "mode": "time",
+            "name": null,
+            "show": true,
+            "values": []
+          },
+          "yaxes": [
+            {
+              "format": "hertz",
+              "logBase": 1,
+              "max": null,
+              "min": null,
+              "show": true
+            },
+            {
+              "format": "short",
+              "logBase": 1,
+              "max": null,
+              "min": null,
+              "show": true
+            }
+          ],
+          "yaxis": {
+            "align": false,
+            "alignLevel": null
+          }
+        },
+        {
+          "aliasColors": {},
+          "bars": false,
+          "dashLength": 10,
+          "dashes": false,
+          "datasource": "$datasource",
+          "decimals": null,
+          "editable": true,
+          "error": false,
+          "fill": 2,
+          "grid": {},
+          "gridPos": {
+            "h": 8,
+            "w": 12,
+            "x": 0,
+            "y": 56
+          },
+          "id": 23,
+          "legend": {
+            "alignAsTable": true,
+            "avg": false,
+            "current": false,
+            "hideEmpty": false,
+            "hideZero": true,
+            "max": false,
+            "min": false,
+            "rightSide": false,
+            "show": true,
+            "total": false,
+            "values": false
+          },
+          "lines": true,
+          "linewidth": 1,
+          "links": [],
+          "nullPointMode": "null",
+          "percentage": false,
+          "pointradius": 5,
+          "points": false,
+          "renderer": "flot",
+          "seriesOverrides": [],
+          "spaceLength": 10,
+          "stack": false,
+          "steppedLine": false,
+          "targets": [
+            {
+              "expr": "rate(synapse_http_server_response_ru_utime_seconds{instance=\"$instance\",job=~\"$job\",index=~\"$index\"}[$bucket_size])+rate(synapse_http_server_response_ru_stime_seconds{instance=\"$instance\",job=~\"$job\",index=~\"$index\"}[$bucket_size])",
+              "format": "time_series",
+              "interval": "",
+              "intervalFactor": 1,
+              "legendFormat": "{{job}}-{{index}} {{method}} {{servlet}} {{tag}}",
+              "refId": "A",
+              "step": 20
+            }
+          ],
+          "thresholds": [
+            {
+              "colorMode": "custom",
+              "fill": true,
+              "fillColor": "rgba(216, 200, 27, 0.27)",
+              "op": "gt",
+              "value": 100,
+              "yaxis": "left"
+            },
+            {
+              "colorMode": "custom",
+              "fill": true,
+              "fillColor": "rgba(234, 112, 112, 0.22)",
+              "op": "gt",
+              "value": 250,
+              "yaxis": "left"
+            }
+          ],
+          "timeFrom": null,
+          "timeShift": null,
+          "title": "Total CPU Usage by Endpoint",
+          "tooltip": {
+            "shared": false,
+            "sort": 0,
+            "value_type": "individual"
+          },
+          "transparent": false,
+          "type": "graph",
+          "xaxis": {
+            "buckets": null,
+            "mode": "time",
+            "name": null,
+            "show": true,
+            "values": []
+          },
+          "yaxes": [
+            {
+              "format": "percentunit",
+              "logBase": 1,
+              "max": null,
+              "min": null,
+              "show": true
+            },
+            {
+              "format": "short",
+              "logBase": 1,
+              "max": null,
+              "min": null,
+              "show": true
+            }
+          ],
+          "yaxis": {
+            "align": false,
+            "alignLevel": null
+          }
+        },
+        {
+          "aliasColors": {},
+          "bars": false,
+          "dashLength": 10,
+          "dashes": false,
+          "datasource": "$datasource",
+          "decimals": null,
+          "editable": true,
+          "error": false,
+          "fill": 2,
+          "grid": {},
+          "gridPos": {
+            "h": 8,
+            "w": 12,
+            "x": 12,
+            "y": 56
+          },
+          "id": 52,
+          "legend": {
+            "alignAsTable": true,
+            "avg": false,
+            "current": false,
+            "hideEmpty": false,
+            "hideZero": true,
+            "max": false,
+            "min": false,
+            "rightSide": false,
+            "show": true,
+            "total": false,
+            "values": false
+          },
+          "lines": true,
+          "linewidth": 1,
+          "links": [],
+          "nullPointMode": "null",
+          "percentage": false,
+          "pointradius": 5,
+          "points": false,
+          "renderer": "flot",
+          "seriesOverrides": [],
+          "spaceLength": 10,
+          "stack": false,
+          "steppedLine": false,
+          "targets": [
+            {
+              "expr": "(rate(synapse_http_server_response_ru_utime_seconds{instance=\"$instance\",job=~\"$job\",index=~\"$index\"}[$bucket_size])+rate(synapse_http_server_response_ru_stime_seconds{instance=\"$instance\",job=~\"$job\",index=~\"$index\"}[$bucket_size])) / rate(synapse_http_server_response_count{instance=\"$instance\",job=~\"$job\",index=~\"$index\"}[$bucket_size])",
+              "format": "time_series",
+              "interval": "",
+              "intervalFactor": 2,
+              "legendFormat": "{{job}}-{{index}} {{method}} {{servlet}} {{tag}}",
+              "refId": "A",
+              "step": 20
+            }
+          ],
+          "thresholds": [
+            {
+              "colorMode": "custom",
+              "fill": true,
+              "fillColor": "rgba(216, 200, 27, 0.27)",
+              "op": "gt",
+              "value": 100
+            },
+            {
+              "colorMode": "custom",
+              "fill": true,
+              "fillColor": "rgba(234, 112, 112, 0.22)",
+              "op": "gt",
+              "value": 250
+            }
+          ],
+          "timeFrom": null,
+          "timeShift": null,
+          "title": "Average CPU Usage by Endpoint",
+          "tooltip": {
+            "shared": false,
+            "sort": 0,
+            "value_type": "individual"
+          },
+          "transparent": false,
+          "type": "graph",
+          "xaxis": {
+            "buckets": null,
+            "mode": "time",
+            "name": null,
+            "show": true,
+            "values": []
+          },
+          "yaxes": [
+            {
+              "format": "s",
+              "logBase": 1,
+              "max": null,
+              "min": null,
+              "show": true
+            },
+            {
+              "format": "short",
+              "logBase": 1,
+              "max": null,
+              "min": null,
+              "show": true
+            }
+          ],
+          "yaxis": {
+            "align": false,
+            "alignLevel": null
+          }
+        },
+        {
+          "aliasColors": {},
+          "bars": false,
+          "dashLength": 10,
+          "dashes": false,
+          "datasource": "$datasource",
+          "editable": true,
+          "error": false,
+          "fill": 1,
+          "grid": {},
+          "gridPos": {
+            "h": 8,
+            "w": 12,
+            "x": 0,
+            "y": 64
+          },
+          "id": 7,
+          "legend": {
+            "alignAsTable": true,
+            "avg": false,
+            "current": false,
+            "hideEmpty": true,
+            "hideZero": true,
+            "max": false,
+            "min": false,
+            "show": true,
+            "total": false,
+            "values": false
+          },
+          "lines": true,
+          "linewidth": 1,
+          "links": [],
+          "nullPointMode": "null",
+          "percentage": false,
+          "pointradius": 5,
+          "points": false,
+          "renderer": "flot",
+          "seriesOverrides": [],
+          "spaceLength": 10,
+          "stack": false,
+          "steppedLine": false,
+          "targets": [
+            {
+              "expr": "rate(synapse_http_server_response_db_txn_duration_seconds{instance=\"$instance\",job=~\"$job\",index=~\"$index\"}[$bucket_size])",
+              "format": "time_series",
+              "interval": "",
+              "intervalFactor": 2,
+              "legendFormat": "{{job}}-{{index}} {{method}} {{servlet}} {{tag}}",
+              "refId": "A",
+              "step": 20
+            }
+          ],
+          "thresholds": [],
+          "timeFrom": null,
+          "timeShift": null,
+          "title": "DB Usage by endpoint",
+          "tooltip": {
+            "shared": false,
+            "sort": 0,
+            "value_type": "cumulative"
+          },
+          "type": "graph",
+          "xaxis": {
+            "buckets": null,
+            "mode": "time",
+            "name": null,
+            "show": true,
+            "values": []
+          },
+          "yaxes": [
+            {
+              "format": "percentunit",
+              "logBase": 1,
+              "max": null,
+              "min": null,
+              "show": true
+            },
+            {
+              "format": "short",
+              "logBase": 1,
+              "max": null,
+              "min": null,
+              "show": true
+            }
+          ],
+          "yaxis": {
+            "align": false,
+            "alignLevel": null
+          }
+        },
+        {
+          "aliasColors": {},
+          "bars": false,
+          "dashLength": 10,
+          "dashes": false,
+          "datasource": "$datasource",
+          "decimals": null,
+          "editable": true,
+          "error": false,
+          "fill": 2,
+          "grid": {},
+          "gridPos": {
+            "h": 8,
+            "w": 12,
+            "x": 12,
+            "y": 64
+          },
+          "id": 47,
+          "legend": {
+            "alignAsTable": true,
+            "avg": true,
+            "current": false,
+            "hideEmpty": false,
+            "hideZero": true,
+            "max": true,
+            "min": false,
+            "rightSide": false,
+            "show": true,
+            "total": false,
+            "values": true
+          },
+          "lines": true,
+          "linewidth": 1,
+          "links": [],
+          "nullPointMode": "null",
+          "percentage": false,
+          "pointradius": 5,
+          "points": false,
+          "renderer": "flot",
+          "seriesOverrides": [],
+          "spaceLength": 10,
+          "stack": false,
+          "steppedLine": false,
+          "targets": [
+            {
+              "expr": "rate(synapse_http_server_response_time_seconds_sum{instance=\"$instance\",job=~\"$job\",index=~\"$index\",tag!=\"incremental_sync\"}[$bucket_size])/rate(synapse_http_server_response_time_seconds_count{instance=\"$instance\",job=~\"$job\",index=~\"$index\",tag!=\"incremental_sync\"}[$bucket_size])",
+              "format": "time_series",
+              "interval": "",
+              "intervalFactor": 2,
+              "legendFormat": "{{job}}-{{index}} {{method}} {{servlet}} {{tag}}",
+              "refId": "A",
+              "step": 20
+            }
+          ],
+          "thresholds": [],
+          "timeFrom": null,
+          "timeShift": null,
+          "title": "Non-sync avg response time",
+          "tooltip": {
+            "shared": false,
+            "sort": 0,
+            "value_type": "individual"
+          },
+          "transparent": false,
+          "type": "graph",
+          "xaxis": {
+            "buckets": null,
+            "mode": "time",
+            "name": null,
+            "show": true,
+            "values": []
+          },
+          "yaxes": [
+            {
+              "format": "s",
+              "logBase": 1,
+              "max": null,
+              "min": null,
+              "show": true
+            },
+            {
+              "format": "short",
+              "logBase": 1,
+              "max": null,
+              "min": null,
+              "show": false
+            }
+          ],
+          "yaxis": {
+            "align": false,
+            "alignLevel": null
+          }
+        },
+        {
+          "aliasColors": {},
+          "bars": false,
+          "dashLength": 10,
+          "dashes": false,
+          "datasource": "${DS_PROMETHEUS}",
+          "fill": 1,
+          "gridPos": {
+            "h": 9,
+            "w": 12,
+            "x": 0,
+            "y": 72
+          },
+          "id": 103,
+          "legend": {
+            "avg": false,
+            "current": false,
+            "max": false,
+            "min": false,
+            "show": true,
+            "total": false,
+            "values": false
+          },
+          "lines": true,
+          "linewidth": 1,
+          "links": [],
+          "nullPointMode": "null",
+          "percentage": false,
+          "pointradius": 5,
+          "points": false,
+          "renderer": "flot",
+          "seriesOverrides": [],
+          "spaceLength": 10,
+          "stack": false,
+          "steppedLine": false,
+          "targets": [
+            {
+              "expr": "topk(10,synapse_http_server_in_flight_requests_count{instance=\"$instance\",job=~\"$job\",index=~\"$index\"})",
+              "format": "time_series",
+              "interval": "",
+              "intervalFactor": 1,
+              "legendFormat": "{{job}}-{{index}} {{method}} {{servlet}}",
+              "refId": "A"
+            }
+          ],
+          "thresholds": [],
+          "timeFrom": null,
+          "timeShift": null,
+          "title": "Requests in flight",
+          "tooltip": {
+            "shared": false,
+            "sort": 0,
+            "value_type": "individual"
+          },
+          "type": "graph",
+          "xaxis": {
+            "buckets": null,
+            "mode": "time",
+            "name": null,
+            "show": true,
+            "values": []
+          },
+          "yaxes": [
+            {
+              "format": "short",
+              "label": null,
+              "logBase": 1,
+              "max": null,
+              "min": null,
+              "show": true
+            },
+            {
+              "format": "short",
+              "label": null,
+              "logBase": 1,
+              "max": null,
+              "min": null,
+              "show": true
+            }
+          ],
+          "yaxis": {
+            "align": false,
+            "alignLevel": null
+          }
+        }
+      ],
+      "repeat": null,
+      "title": "Requests",
+      "type": "row"
+    },
+    {
+      "collapsed": true,
+      "gridPos": {
+        "h": 1,
+        "w": 24,
+        "x": 0,
+        "y": 20
+      },
+      "id": 97,
+      "panels": [
+        {
+          "aliasColors": {},
+          "bars": false,
+          "dashLength": 10,
+          "dashes": false,
+          "datasource": "${DS_PROMETHEUS}",
+          "fill": 1,
+          "gridPos": {
+            "h": 9,
+            "w": 12,
+            "x": 0,
+            "y": 23
+          },
+          "id": 99,
+          "legend": {
+            "avg": false,
+            "current": false,
+            "max": false,
+            "min": false,
+            "show": true,
+            "total": false,
+            "values": false
+          },
+          "lines": true,
+          "linewidth": 1,
+          "links": [],
+          "nullPointMode": "null",
+          "percentage": false,
+          "pointradius": 5,
+          "points": false,
+          "renderer": "flot",
+          "seriesOverrides": [],
+          "spaceLength": 10,
+          "stack": false,
+          "steppedLine": false,
+          "targets": [
+            {
+              "expr": "rate(synapse_background_process_ru_utime_seconds{instance=\"$instance\",job=~\"$job\",index=~\"$index\"}[$bucket_size])+rate(synapse_background_process_ru_stime_seconds{instance=\"$instance\",job=~\"$job\",index=~\"$index\"}[$bucket_size])",
+              "format": "time_series",
+              "interval": "",
+              "intervalFactor": 1,
+              "legendFormat": "{{job}}-{{index}} {{name}}",
+              "refId": "A"
+            }
+          ],
+          "thresholds": [],
+          "timeFrom": null,
+          "timeShift": null,
+          "title": "CPU usage by background jobs",
+          "tooltip": {
+            "shared": true,
+            "sort": 0,
+            "value_type": "individual"
+          },
+          "type": "graph",
+          "xaxis": {
+            "buckets": null,
+            "mode": "time",
+            "name": null,
+            "show": true,
+            "values": []
+          },
+          "yaxes": [
+            {
+              "format": "percentunit",
+              "label": null,
+              "logBase": 1,
+              "max": null,
+              "min": null,
+              "show": true
+            },
+            {
+              "format": "short",
+              "label": null,
+              "logBase": 1,
+              "max": null,
+              "min": null,
+              "show": true
+            }
+          ],
+          "yaxis": {
+            "align": false,
+            "alignLevel": null
+          }
+        },
+        {
+          "aliasColors": {},
+          "bars": false,
+          "dashLength": 10,
+          "dashes": false,
+          "datasource": "${DS_PROMETHEUS}",
+          "fill": 1,
+          "gridPos": {
+            "h": 9,
+            "w": 12,
+            "x": 12,
+            "y": 23
+          },
+          "id": 101,
+          "legend": {
+            "avg": false,
+            "current": false,
+            "max": false,
+            "min": false,
+            "show": true,
+            "total": false,
+            "values": false
+          },
+          "lines": true,
+          "linewidth": 1,
+          "links": [],
+          "nullPointMode": "null",
+          "percentage": false,
+          "pointradius": 5,
+          "points": false,
+          "renderer": "flot",
+          "seriesOverrides": [],
+          "spaceLength": 10,
+          "stack": false,
+          "steppedLine": false,
+          "targets": [
+            {
+              "expr": "rate(synapse_background_process_db_txn_duration_seconds{instance=\"$instance\",job=~\"$job\",index=~\"$index\"}[$bucket_size])",
+              "format": "time_series",
+              "intervalFactor": 1,
+              "legendFormat": "{{job}}-{{index}} {{name}}",
+              "refId": "A"
+            }
+          ],
+          "thresholds": [],
+          "timeFrom": null,
+          "timeShift": null,
+          "title": "DB usage by background jobs",
+          "tooltip": {
+            "shared": true,
+            "sort": 0,
+            "value_type": "individual"
+          },
+          "type": "graph",
+          "xaxis": {
+            "buckets": null,
+            "mode": "time",
+            "name": null,
+            "show": true,
+            "values": []
+          },
+          "yaxes": [
+            {
+              "format": "percentunit",
+              "label": null,
+              "logBase": 1,
+              "max": null,
+              "min": null,
+              "show": true
+            },
+            {
+              "format": "short",
+              "label": null,
+              "logBase": 1,
+              "max": null,
+              "min": null,
+              "show": true
+            }
+          ],
+          "yaxis": {
+            "align": false,
+            "alignLevel": null
+          }
+        }
+      ],
+      "title": "Background jobs",
+      "type": "row"
+    },
+    {
+      "collapsed": true,
+      "gridPos": {
+        "h": 1,
+        "w": 24,
+        "x": 0,
+        "y": 21
+      },
+      "id": 81,
+      "panels": [
+        {
+          "aliasColors": {},
+          "bars": false,
+          "dashLength": 10,
+          "dashes": false,
+          "datasource": "${DS_PROMETHEUS}",
+          "fill": 1,
+          "gridPos": {
+            "h": 9,
+            "w": 12,
+            "x": 0,
+            "y": 25
+          },
+          "id": 79,
+          "legend": {
+            "avg": false,
+            "current": false,
+            "max": false,
+            "min": false,
+            "show": true,
+            "total": false,
+            "values": false
+          },
+          "lines": true,
+          "linewidth": 1,
+          "links": [],
+          "nullPointMode": "null",
+          "percentage": false,
+          "pointradius": 5,
+          "points": false,
+          "renderer": "flot",
+          "seriesOverrides": [],
+          "spaceLength": 10,
+          "stack": false,
+          "steppedLine": false,
+          "targets": [
+            {
+              "expr": "rate(synapse_federation_client_sent_transactions{instance=\"$instance\", job=~\"$job\", index=~\"$index\"}[$bucket_size])",
+              "format": "time_series",
+              "intervalFactor": 1,
+              "legendFormat": "txn rate",
+              "refId": "A"
+            }
+          ],
+          "thresholds": [],
+          "timeFrom": null,
+          "timeShift": null,
+          "title": "Outgoing federation transaction rate",
+          "tooltip": {
+            "shared": true,
+            "sort": 0,
+            "value_type": "individual"
+          },
+          "type": "graph",
+          "xaxis": {
+            "buckets": null,
+            "mode": "time",
+            "name": null,
+            "show": true,
+            "values": []
+          },
+          "yaxes": [
+            {
+              "format": "hertz",
+              "label": null,
+              "logBase": 1,
+              "max": null,
+              "min": null,
+              "show": true
+            },
+            {
+              "format": "short",
+              "label": null,
+              "logBase": 1,
+              "max": null,
+              "min": null,
+              "show": true
+            }
+          ],
+          "yaxis": {
+            "align": false,
+            "alignLevel": null
+          }
+        },
+        {
+          "aliasColors": {},
+          "bars": false,
+          "dashLength": 10,
+          "dashes": false,
+          "datasource": "${DS_PROMETHEUS}",
+          "fill": 1,
+          "gridPos": {
+            "h": 9,
+            "w": 12,
+            "x": 12,
+            "y": 25
+          },
+          "id": 83,
+          "legend": {
+            "avg": false,
+            "current": false,
+            "max": false,
+            "min": false,
+            "show": true,
+            "total": false,
+            "values": false
+          },
+          "lines": true,
+          "linewidth": 1,
+          "links": [],
+          "nullPointMode": "null",
+          "percentage": false,
+          "pointradius": 5,
+          "points": false,
+          "renderer": "flot",
+          "seriesOverrides": [],
+          "spaceLength": 10,
+          "stack": false,
+          "steppedLine": false,
+          "targets": [
+            {
+              "expr": "rate(synapse_federation_server_received_pdus{job=~\"$job\",index=~\"$index\",instance=\"$instance\"}[$bucket_size])",
+              "format": "time_series",
+              "intervalFactor": 1,
+              "legendFormat": "pdus",
+              "refId": "A"
+            },
+            {
+              "expr": "rate(synapse_federation_server_received_edus{job=~\"$job\",index=~\"$index\",instance=\"$instance\"}[$bucket_size])",
+              "format": "time_series",
+              "intervalFactor": 1,
+              "legendFormat": "edus",
+              "refId": "B"
+            }
+          ],
+          "thresholds": [],
+          "timeFrom": null,
+          "timeShift": null,
+          "title": "Incoming PDU/EDU rate",
+          "tooltip": {
+            "shared": true,
+            "sort": 0,
+            "value_type": "individual"
+          },
+          "type": "graph",
+          "xaxis": {
+            "buckets": null,
+            "mode": "time",
+            "name": null,
+            "show": true,
+            "values": []
+          },
+          "yaxes": [
+            {
+              "format": "hertz",
+              "label": null,
+              "logBase": 1,
+              "max": null,
+              "min": null,
+              "show": true
+            },
+            {
+              "format": "short",
+              "label": null,
+              "logBase": 1,
+              "max": null,
+              "min": null,
+              "show": true
+            }
+          ],
+          "yaxis": {
+            "align": false,
+            "alignLevel": null
+          }
+        }
+      ],
+      "title": "Federation",
+      "type": "row"
+    },
+    {
+      "collapsed": true,
+      "gridPos": {
+        "h": 1,
+        "w": 24,
+        "x": 0,
+        "y": 22
+      },
+      "id": 60,
+      "panels": [
+        {
+          "aliasColors": {},
+          "bars": false,
+          "dashLength": 10,
+          "dashes": false,
+          "datasource": "$datasource",
+          "fill": 1,
+          "gridPos": {
+            "h": 7,
+            "w": 12,
+            "x": 0,
+            "y": 23
+          },
+          "id": 51,
+          "legend": {
+            "avg": false,
+            "current": false,
+            "max": false,
+            "min": false,
+            "show": true,
+            "total": false,
+            "values": false
+          },
+          "lines": true,
+          "linewidth": 1,
+          "links": [],
+          "nullPointMode": "null",
+          "percentage": false,
+          "pointradius": 5,
+          "points": false,
+          "renderer": "flot",
+          "seriesOverrides": [],
+          "spaceLength": 10,
+          "stack": false,
+          "steppedLine": false,
+          "targets": [
+            {
+              "expr": "rate(synapse_push_httppusher_http_pushes_processed{instance=\"$instance\",job=~\"$job\",index=~\"$index\"}[$bucket_size])",
+              "format": "time_series",
+              "interval": "",
+              "intervalFactor": 2,
+              "legendFormat": "processed {{job}}",
+              "refId": "A",
+              "step": 20
+            },
+            {
+              "expr": "rate(synapse_push_httppusher_http_pushes_failed{instance=\"$instance\",job=~\"$job\",index=~\"$index\"}[$bucket_size])",
+              "format": "time_series",
+              "intervalFactor": 2,
+              "legendFormat": "failed {{job}}",
+              "refId": "B",
+              "step": 20
+            }
+          ],
+          "thresholds": [],
+          "timeFrom": null,
+          "timeShift": null,
+          "title": "HTTP Push rate",
+          "tooltip": {
+            "shared": true,
+            "sort": 0,
+            "value_type": "individual"
+          },
+          "type": "graph",
+          "xaxis": {
+            "buckets": null,
+            "mode": "time",
+            "name": null,
+            "show": true,
+            "values": []
+          },
+          "yaxes": [
+            {
+              "format": "hertz",
+              "label": null,
+              "logBase": 1,
+              "max": null,
+              "min": null,
+              "show": true
+            },
+            {
+              "format": "short",
+              "label": null,
+              "logBase": 1,
+              "max": null,
+              "min": null,
+              "show": true
+            }
+          ],
+          "yaxis": {
+            "align": false,
+            "alignLevel": null
+          }
+        }
+      ],
+      "repeat": null,
+      "title": "Pushes",
+      "type": "row"
+    },
+    {
+      "collapsed": true,
+      "gridPos": {
+        "h": 1,
+        "w": 24,
+        "x": 0,
+        "y": 23
+      },
+      "id": 58,
+      "panels": [
+        {
+          "aliasColors": {},
+          "bars": false,
+          "dashLength": 10,
+          "dashes": false,
+          "datasource": "$datasource",
+          "editable": true,
+          "error": false,
+          "fill": 0,
+          "grid": {},
+          "gridPos": {
+            "h": 7,
+            "w": 12,
+            "x": 0,
+            "y": 25
+          },
+          "id": 10,
+          "legend": {
+            "avg": false,
+            "current": false,
+            "hideEmpty": true,
+            "hideZero": true,
+            "max": false,
+            "min": false,
+            "show": true,
+            "total": false,
+            "values": false
+          },
+          "lines": true,
+          "linewidth": 2,
+          "links": [],
+          "nullPointMode": "null",
+          "percentage": false,
+          "pointradius": 5,
+          "points": false,
+          "renderer": "flot",
+          "seriesOverrides": [],
+          "spaceLength": 10,
+          "stack": false,
+          "steppedLine": false,
+          "targets": [
+            {
+              "expr": "topk(10, rate(synapse_storage_transaction_time_count{instance=\"$instance\",job=~\"$job\",index=~\"$index\"}[$bucket_size]))",
+              "format": "time_series",
+              "interval": "",
+              "intervalFactor": 2,
+              "legendFormat": "{{job}}-{{index}} {{desc}}",
+              "refId": "A",
+              "step": 20
+            }
+          ],
+          "thresholds": [],
+          "timeFrom": null,
+          "timeShift": null,
+          "title": "Top DB transactions by txn rate",
+          "tooltip": {
+            "shared": false,
+            "sort": 0,
+            "value_type": "cumulative"
+          },
+          "type": "graph",
+          "xaxis": {
+            "buckets": null,
+            "mode": "time",
+            "name": null,
+            "show": true,
+            "values": []
+          },
+          "yaxes": [
+            {
+              "format": "hertz",
+              "logBase": 1,
+              "max": null,
+              "min": 0,
+              "show": true
+            },
+            {
+              "format": "short",
+              "logBase": 1,
+              "max": null,
+              "min": null,
+              "show": true
+            }
+          ],
+          "yaxis": {
+            "align": false,
+            "alignLevel": null
+          }
+        },
+        {
+          "aliasColors": {},
+          "bars": false,
+          "dashLength": 10,
+          "dashes": false,
+          "datasource": "$datasource",
+          "editable": true,
+          "error": false,
+          "fill": 1,
+          "grid": {},
+          "gridPos": {
+            "h": 7,
+            "w": 12,
+            "x": 12,
+            "y": 25
+          },
+          "id": 11,
+          "legend": {
+            "avg": false,
+            "current": false,
+            "hideEmpty": true,
+            "hideZero": true,
+            "max": false,
+            "min": false,
+            "show": true,
+            "total": false,
+            "values": false
+          },
+          "lines": true,
+          "linewidth": 1,
+          "links": [],
+          "nullPointMode": "null",
+          "percentage": false,
+          "pointradius": 5,
+          "points": false,
+          "renderer": "flot",
+          "seriesOverrides": [],
+          "spaceLength": 10,
+          "stack": false,
+          "steppedLine": true,
+          "targets": [
+            {
+              "expr": "topk(5, rate(synapse_storage_transaction_time_sum{instance=\"$instance\",job=~\"$job\",index=~\"$index\"}[$bucket_size]))",
+              "format": "time_series",
+              "instant": false,
+              "interval": "",
+              "intervalFactor": 1,
+              "legendFormat": "{{job}}-{{index}} {{desc}}",
+              "refId": "A",
+              "step": 20
+            }
+          ],
+          "thresholds": [],
+          "timeFrom": null,
+          "timeShift": null,
+          "title": "Top DB transactions by total txn time",
+          "tooltip": {
+            "shared": false,
+            "sort": 0,
+            "value_type": "cumulative"
+          },
+          "type": "graph",
+          "xaxis": {
+            "buckets": null,
+            "mode": "time",
+            "name": null,
+            "show": true,
+            "values": []
+          },
+          "yaxes": [
+            {
+              "format": "percentunit",
+              "logBase": 1,
+              "max": null,
+              "min": null,
+              "show": true
+            },
+            {
+              "format": "short",
+              "logBase": 1,
+              "max": null,
+              "min": null,
+              "show": true
+            }
+          ],
+          "yaxis": {
+            "align": false,
+            "alignLevel": null
+          }
+        }
+      ],
+      "repeat": null,
+      "title": "Database",
+      "type": "row"
+    },
+    {
+      "collapsed": true,
+      "gridPos": {
+        "h": 1,
+        "w": 24,
+        "x": 0,
+        "y": 24
+      },
+      "id": 59,
+      "panels": [
+        {
+          "aliasColors": {},
+          "bars": false,
+          "dashLength": 10,
+          "dashes": false,
+          "datasource": "$datasource",
+          "editable": true,
+          "error": false,
+          "fill": 1,
+          "grid": {},
+          "gridPos": {
+            "h": 13,
+            "w": 12,
+            "x": 0,
+            "y": 17
+          },
+          "id": 12,
+          "legend": {
+            "alignAsTable": true,
+            "avg": false,
+            "current": false,
+            "max": false,
+            "min": false,
+            "show": true,
+            "total": false,
+            "values": false
+          },
+          "lines": true,
+          "linewidth": 2,
+          "links": [],
+          "nullPointMode": "null",
+          "percentage": false,
+          "pointradius": 5,
+          "points": false,
+          "renderer": "flot",
+          "seriesOverrides": [],
+          "spaceLength": 10,
+          "stack": false,
+          "steppedLine": false,
+          "targets": [
+            {
+              "expr": "rate(synapse_util_metrics_block_ru_utime_seconds{instance=\"$instance\",job=~\"$job\",index=~\"$index\",block_name!=\"wrapped_request_handler\"}[$bucket_size]) + rate(synapse_util_metrics_block_ru_stime_seconds[$bucket_size])",
+              "format": "time_series",
+              "interval": "",
+              "intervalFactor": 2,
+              "legendFormat": "{{job}}-{{index}} {{block_name}}",
+              "refId": "A",
+              "step": 20
+            }
+          ],
+          "thresholds": [],
+          "timeFrom": null,
+          "timeShift": null,
+          "title": "Total CPU Usage by Block",
+          "tooltip": {
+            "shared": false,
+            "sort": 0,
+            "value_type": "cumulative"
+          },
+          "type": "graph",
+          "xaxis": {
+            "buckets": null,
+            "mode": "time",
+            "name": null,
+            "show": true,
+            "values": []
+          },
+          "yaxes": [
+            {
+              "format": "percentunit",
+              "logBase": 1,
+              "max": null,
+              "min": null,
+              "show": true
+            },
+            {
+              "format": "short",
+              "logBase": 1,
+              "max": null,
+              "min": null,
+              "show": true
+            }
+          ],
+          "yaxis": {
+            "align": false,
+            "alignLevel": null
+          }
+        },
+        {
+          "aliasColors": {},
+          "bars": false,
+          "dashLength": 10,
+          "dashes": false,
+          "datasource": "$datasource",
+          "editable": true,
+          "error": false,
+          "fill": 1,
+          "grid": {},
+          "gridPos": {
+            "h": 13,
+            "w": 12,
+            "x": 12,
+            "y": 17
+          },
+          "id": 26,
+          "legend": {
+            "alignAsTable": true,
+            "avg": false,
+            "current": false,
+            "max": false,
+            "min": false,
+            "show": true,
+            "total": false,
+            "values": false
+          },
+          "lines": true,
+          "linewidth": 2,
+          "links": [],
+          "nullPointMode": "null",
+          "percentage": false,
+          "pointradius": 5,
+          "points": false,
+          "renderer": "flot",
+          "seriesOverrides": [],
+          "spaceLength": 10,
+          "stack": false,
+          "steppedLine": false,
+          "targets": [
+            {
+              "expr": "(rate(synapse_util_metrics_block_ru_utime_seconds{instance=\"$instance\",job=~\"$job\",index=~\"$index\"}[$bucket_size]) + rate(synapse_util_metrics_block_ru_stime_seconds[$bucket_size])) / rate(synapse_util_metrics_block_count[$bucket_size])",
+              "format": "time_series",
+              "interval": "",
+              "intervalFactor": 2,
+              "legendFormat": "{{job}}-{{index}} {{block_name}}",
+              "refId": "A",
+              "step": 20
+            }
+          ],
+          "thresholds": [],
+          "timeFrom": null,
+          "timeShift": null,
+          "title": "Average CPU Time per Block",
+          "tooltip": {
+            "shared": false,
+            "sort": 0,
+            "value_type": "cumulative"
+          },
+          "type": "graph",
+          "xaxis": {
+            "buckets": null,
+            "mode": "time",
+            "name": null,
+            "show": true,
+            "values": []
+          },
+          "yaxes": [
+            {
+              "format": "ms",
+              "logBase": 1,
+              "max": null,
+              "min": null,
+              "show": true
+            },
+            {
+              "format": "short",
+              "logBase": 1,
+              "max": null,
+              "min": null,
+              "show": true
+            }
+          ],
+          "yaxis": {
+            "align": false,
+            "alignLevel": null
+          }
+        },
+        {
+          "aliasColors": {},
+          "bars": false,
+          "dashLength": 10,
+          "dashes": false,
+          "datasource": "$datasource",
+          "editable": true,
+          "error": false,
+          "fill": 1,
+          "grid": {},
+          "gridPos": {
+            "h": 13,
+            "w": 12,
+            "x": 0,
+            "y": 30
+          },
+          "id": 13,
+          "legend": {
+            "alignAsTable": true,
+            "avg": false,
+            "current": false,
+            "max": false,
+            "min": false,
+            "show": true,
+            "total": false,
+            "values": false
+          },
+          "lines": true,
+          "linewidth": 2,
+          "links": [],
+          "nullPointMode": "null",
+          "percentage": false,
+          "pointradius": 5,
+          "points": false,
+          "renderer": "flot",
+          "seriesOverrides": [],
+          "spaceLength": 10,
+          "stack": false,
+          "steppedLine": false,
+          "targets": [
+            {
+              "expr": "rate(synapse_util_metrics_block_db_txn_duration_seconds{instance=\"$instance\",job=~\"$job\",index=~\"$index\",block_name!=\"wrapped_request_handler\"}[$bucket_size])",
+              "format": "time_series",
+              "interval": "",
+              "intervalFactor": 2,
+              "legendFormat": "{{job}} {{block_name}}",
+              "refId": "A",
+              "step": 20
+            }
+          ],
+          "thresholds": [],
+          "timeFrom": null,
+          "timeShift": null,
+          "title": "Total DB Usage by Block",
+          "tooltip": {
+            "shared": false,
+            "sort": 0,
+            "value_type": "cumulative"
+          },
+          "type": "graph",
+          "xaxis": {
+            "buckets": null,
+            "mode": "time",
+            "name": null,
+            "show": true,
+            "values": []
+          },
+          "yaxes": [
+            {
+              "format": "percentunit",
+              "logBase": 1,
+              "max": null,
+              "min": 0,
+              "show": true
+            },
+            {
+              "format": "short",
+              "logBase": 1,
+              "max": null,
+              "min": null,
+              "show": true
+            }
+          ],
+          "yaxis": {
+            "align": false,
+            "alignLevel": null
+          }
+        },
+        {
+          "aliasColors": {},
+          "bars": false,
+          "dashLength": 10,
+          "dashes": false,
+          "datasource": "$datasource",
+          "editable": true,
+          "error": false,
+          "fill": 1,
+          "grid": {},
+          "gridPos": {
+            "h": 13,
+            "w": 12,
+            "x": 12,
+            "y": 30
+          },
+          "id": 27,
+          "legend": {
+            "alignAsTable": true,
+            "avg": false,
+            "current": false,
+            "max": false,
+            "min": false,
+            "show": true,
+            "total": false,
+            "values": false
+          },
+          "lines": true,
+          "linewidth": 2,
+          "links": [],
+          "nullPointMode": "null",
+          "percentage": false,
+          "pointradius": 5,
+          "points": false,
+          "renderer": "flot",
+          "seriesOverrides": [],
+          "spaceLength": 10,
+          "stack": false,
+          "steppedLine": false,
+          "targets": [
+            {
+              "expr": "rate(synapse_util_metrics_block_db_txn_duration_seconds{instance=\"$instance\",job=~\"$job\",index=~\"$index\"}[$bucket_size]) / rate(synapse_util_metrics_block_db_txn_count{instance=\"$instance\",job=~\"$job\",index=~\"$index\"}[$bucket_size])",
+              "format": "time_series",
+              "interval": "",
+              "intervalFactor": 2,
+              "legendFormat": "{{job}}-{{index}} {{block_name}}",
+              "refId": "A",
+              "step": 20
+            }
+          ],
+          "thresholds": [],
+          "timeFrom": null,
+          "timeShift": null,
+          "title": "Average Database Time per Block",
+          "tooltip": {
+            "shared": false,
+            "sort": 0,
+            "value_type": "cumulative"
+          },
+          "type": "graph",
+          "xaxis": {
+            "buckets": null,
+            "mode": "time",
+            "name": null,
+            "show": true,
+            "values": []
+          },
+          "yaxes": [
+            {
+              "format": "ms",
+              "logBase": 1,
+              "max": null,
+              "min": null,
+              "show": true
+            },
+            {
+              "format": "short",
+              "logBase": 1,
+              "max": null,
+              "min": null,
+              "show": true
+            }
+          ],
+          "yaxis": {
+            "align": false,
+            "alignLevel": null
+          }
+        },
+        {
+          "aliasColors": {},
+          "bars": false,
+          "dashLength": 10,
+          "dashes": false,
+          "datasource": "$datasource",
+          "editable": true,
+          "error": false,
+          "fill": 1,
+          "grid": {},
+          "gridPos": {
+            "h": 13,
+            "w": 12,
+            "x": 0,
+            "y": 43
+          },
+          "id": 28,
+          "legend": {
+            "avg": false,
+            "current": false,
+            "max": false,
+            "min": false,
+            "show": false,
+            "total": false,
+            "values": false
+          },
+          "lines": true,
+          "linewidth": 2,
+          "links": [],
+          "nullPointMode": "null",
+          "percentage": false,
+          "pointradius": 5,
+          "points": false,
+          "renderer": "flot",
+          "seriesOverrides": [],
+          "spaceLength": 10,
+          "stack": false,
+          "steppedLine": false,
+          "targets": [
+            {
+              "expr": "rate(synapse_util_metrics_block_db_txn_duration_seconds{instance=\"$instance\",job=~\"$job\",index=~\"$index\"}[$bucket_size]) / rate(synapse_util_metrics_block_db_txn_count{instance=\"$instance\",job=~\"$job\",index=~\"$index\"}[$bucket_size])",
+              "format": "time_series",
+              "interval": "",
+              "intervalFactor": 2,
+              "legendFormat": "{{job}}-{{index}} {{block_name}}",
+              "refId": "A",
+              "step": 20
+            }
+          ],
+          "thresholds": [],
+          "timeFrom": null,
+          "timeShift": null,
+          "title": "Average Transactions per Block",
+          "tooltip": {
+            "shared": false,
+            "sort": 0,
+            "value_type": "cumulative"
+          },
+          "type": "graph",
+          "xaxis": {
+            "buckets": null,
+            "mode": "time",
+            "name": null,
+            "show": true,
+            "values": []
+          },
+          "yaxes": [
+            {
+              "format": "none",
+              "logBase": 1,
+              "max": null,
+              "min": null,
+              "show": true
+            },
+            {
+              "format": "short",
+              "logBase": 1,
+              "max": null,
+              "min": null,
+              "show": true
+            }
+          ],
+          "yaxis": {
+            "align": false,
+            "alignLevel": null
+          }
+        },
+        {
+          "aliasColors": {},
+          "bars": false,
+          "dashLength": 10,
+          "dashes": false,
+          "datasource": "$datasource",
+          "editable": true,
+          "error": false,
+          "fill": 1,
+          "grid": {},
+          "gridPos": {
+            "h": 13,
+            "w": 12,
+            "x": 12,
+            "y": 43
+          },
+          "id": 25,
+          "legend": {
+            "avg": false,
+            "current": false,
+            "max": false,
+            "min": false,
+            "show": false,
+            "total": false,
+            "values": false
+          },
+          "lines": true,
+          "linewidth": 2,
+          "links": [],
+          "nullPointMode": "null",
+          "percentage": false,
+          "pointradius": 5,
+          "points": false,
+          "renderer": "flot",
+          "seriesOverrides": [],
+          "spaceLength": 10,
+          "stack": false,
+          "steppedLine": false,
+          "targets": [
+            {
+              "expr": "rate(synapse_util_metrics_block_time_seconds{instance=\"$instance\",job=~\"$job\",index=~\"$index\"}[$bucket_size]) / rate(synapse_util_metrics_block_count[$bucket_size])",
+              "format": "time_series",
+              "interval": "",
+              "intervalFactor": 2,
+              "legendFormat": "{{job}}-{{index}} {{block_name}}",
+              "refId": "A",
+              "step": 20
+            }
+          ],
+          "thresholds": [],
+          "timeFrom": null,
+          "timeShift": null,
+          "title": "Average Wallclock Time per Block",
+          "tooltip": {
+            "shared": false,
+            "sort": 0,
+            "value_type": "cumulative"
+          },
+          "type": "graph",
+          "xaxis": {
+            "buckets": null,
+            "mode": "time",
+            "name": null,
+            "show": true,
+            "values": []
+          },
+          "yaxes": [
+            {
+              "format": "ms",
+              "logBase": 1,
+              "max": null,
+              "min": null,
+              "show": true
+            },
+            {
+              "format": "short",
+              "logBase": 1,
+              "max": null,
+              "min": null,
+              "show": true
+            }
+          ],
+          "yaxis": {
+            "align": false,
+            "alignLevel": null
+          }
+        }
+      ],
+      "repeat": null,
+      "title": "Per-block metrics",
+      "type": "row"
+    },
+    {
+      "collapsed": true,
+      "gridPos": {
+        "h": 1,
+        "w": 24,
+        "x": 0,
+        "y": 25
+      },
+      "id": 61,
+      "panels": [
+        {
+          "aliasColors": {},
+          "bars": false,
+          "dashLength": 10,
+          "dashes": false,
+          "datasource": "$datasource",
+          "decimals": 2,
+          "editable": true,
+          "error": false,
+          "fill": 0,
+          "grid": {},
+          "gridPos": {
+            "h": 10,
+            "w": 12,
+            "x": 0,
+            "y": 55
+          },
+          "id": 1,
+          "legend": {
+            "alignAsTable": true,
+            "avg": false,
+            "current": false,
+            "hideEmpty": true,
+            "hideZero": false,
+            "max": false,
+            "min": false,
+            "show": true,
+            "total": false,
+            "values": false
+          },
+          "lines": true,
+          "linewidth": 2,
+          "links": [],
+          "nullPointMode": "null",
+          "percentage": false,
+          "pointradius": 5,
+          "points": false,
+          "renderer": "flot",
+          "seriesOverrides": [],
+          "spaceLength": 10,
+          "stack": false,
+          "steppedLine": false,
+          "targets": [
+            {
+              "expr": "rate(synapse_util_caches_cache:hits{job=~\"$job\",index=~\"$index\",instance=\"$instance\"}[$bucket_size])/rate(synapse_util_caches_cache:total{job=~\"$job\",index=~\"$index\",instance=\"$instance\"}[$bucket_size])",
+              "format": "time_series",
+              "intervalFactor": 2,
+              "legendFormat": "{{name}} {{job}}-{{index}}",
+              "refId": "A",
+              "step": 20
+            }
+          ],
+          "thresholds": [],
+          "timeFrom": null,
+          "timeShift": null,
+          "title": "Cache Hit Ratio",
+          "tooltip": {
+            "msResolution": true,
+            "shared": false,
+            "sort": 0,
+            "value_type": "cumulative"
+          },
+          "type": "graph",
+          "xaxis": {
+            "buckets": null,
+            "mode": "time",
+            "name": null,
+            "show": true,
+            "values": []
+          },
+          "yaxes": [
+            {
+              "decimals": null,
+              "format": "percentunit",
+              "label": "",
+              "logBase": 1,
+              "max": "1",
+              "min": 0,
+              "show": true
+            },
+            {
+              "format": "short",
+              "logBase": 1,
+              "max": null,
+              "min": null,
+              "show": false
+            }
+          ],
+          "yaxis": {
+            "align": false,
+            "alignLevel": null
+          }
+        },
+        {
+          "aliasColors": {},
+          "bars": false,
+          "dashLength": 10,
+          "dashes": false,
+          "datasource": "$datasource",
+          "editable": true,
+          "error": false,
+          "fill": 1,
+          "grid": {},
+          "gridPos": {
+            "h": 10,
+            "w": 12,
+            "x": 12,
+            "y": 55
+          },
+          "id": 8,
+          "legend": {
+            "alignAsTable": true,
+            "avg": false,
+            "current": false,
+            "hideZero": false,
+            "max": false,
+            "min": false,
+            "show": true,
+            "total": false,
+            "values": false
+          },
+          "lines": true,
+          "linewidth": 2,
+          "links": [],
+          "nullPointMode": "connected",
+          "percentage": false,
+          "pointradius": 5,
+          "points": false,
+          "renderer": "flot",
+          "seriesOverrides": [],
+          "spaceLength": 10,
+          "stack": false,
+          "steppedLine": false,
+          "targets": [
+            {
+              "expr": "synapse_util_caches_cache:size{instance=\"$instance\",job=~\"$job\",index=~\"$index\"}",
+              "format": "time_series",
+              "hide": false,
+              "interval": "",
+              "intervalFactor": 2,
+              "legendFormat": "{{name}} {{job}}-{{index}}",
+              "refId": "A",
+              "step": 20
+            }
+          ],
+          "thresholds": [],
+          "timeFrom": null,
+          "timeShift": null,
+          "title": "Cache Size",
+          "tooltip": {
+            "shared": false,
+            "sort": 0,
+            "value_type": "cumulative"
+          },
+          "type": "graph",
+          "xaxis": {
+            "buckets": null,
+            "mode": "time",
+            "name": null,
+            "show": true,
+            "values": []
+          },
+          "yaxes": [
+            {
+              "format": "short",
+              "logBase": 1,
+              "max": null,
+              "min": 0,
+              "show": true
+            },
+            {
+              "format": "short",
+              "logBase": 1,
+              "max": null,
+              "min": null,
+              "show": true
+            }
+          ],
+          "yaxis": {
+            "align": false,
+            "alignLevel": null
+          }
+        },
+        {
+          "aliasColors": {},
+          "bars": false,
+          "dashLength": 10,
+          "dashes": false,
+          "datasource": "$datasource",
+          "editable": true,
+          "error": false,
+          "fill": 1,
+          "grid": {},
+          "gridPos": {
+            "h": 10,
+            "w": 12,
+            "x": 0,
+            "y": 65
+          },
+          "id": 38,
+          "legend": {
+            "alignAsTable": true,
+            "avg": false,
+            "current": false,
+            "hideZero": false,
+            "max": false,
+            "min": false,
+            "show": true,
+            "total": false,
+            "values": false
+          },
+          "lines": true,
+          "linewidth": 2,
+          "links": [],
+          "nullPointMode": "connected",
+          "percentage": false,
+          "pointradius": 5,
+          "points": false,
+          "renderer": "flot",
+          "seriesOverrides": [],
+          "spaceLength": 10,
+          "stack": false,
+          "steppedLine": false,
+          "targets": [
+            {
+              "expr": "rate(synapse_util_caches_cache:total{job=~\"$job\",index=~\"$index\",instance=\"$instance\"}[$bucket_size])",
+              "format": "time_series",
+              "interval": "",
+              "intervalFactor": 2,
+              "legendFormat": "{{name}} {{job}}-{{index}}",
+              "refId": "A",
+              "step": 20
+            }
+          ],
+          "thresholds": [],
+          "timeFrom": null,
+          "timeShift": null,
+          "title": "Cache request rate",
+          "tooltip": {
+            "shared": false,
+            "sort": 0,
+            "value_type": "cumulative"
+          },
+          "type": "graph",
+          "xaxis": {
+            "buckets": null,
+            "mode": "time",
+            "name": null,
+            "show": true,
+            "values": []
+          },
+          "yaxes": [
+            {
+              "format": "rps",
+              "logBase": 1,
+              "max": null,
+              "min": 0,
+              "show": true
+            },
+            {
+              "format": "short",
+              "logBase": 1,
+              "max": null,
+              "min": null,
+              "show": true
+            }
+          ],
+          "yaxis": {
+            "align": false,
+            "alignLevel": null
+          }
+        },
+        {
+          "aliasColors": {},
+          "bars": false,
+          "dashLength": 10,
+          "dashes": false,
+          "datasource": "$datasource",
+          "fill": 1,
+          "gridPos": {
+            "h": 10,
+            "w": 12,
+            "x": 12,
+            "y": 65
+          },
+          "id": 39,
+          "legend": {
+            "alignAsTable": true,
+            "avg": false,
+            "current": false,
+            "max": false,
+            "min": false,
+            "show": true,
+            "total": false,
+            "values": false
+          },
+          "lines": true,
+          "linewidth": 1,
+          "links": [],
+          "nullPointMode": "null",
+          "percentage": false,
+          "pointradius": 5,
+          "points": false,
+          "renderer": "flot",
+          "seriesOverrides": [],
+          "spaceLength": 10,
+          "stack": false,
+          "steppedLine": false,
+          "targets": [
+            {
+              "expr": "topk(10, rate(synapse_util_caches_cache:total{job=\"$job\",index=~\"$index\",instance=\"$instance\"}[$bucket_size]) - rate(synapse_util_caches_cache:hits{job=\"$job\",instance=\"$instance\"}[$bucket_size]))",
+              "format": "time_series",
+              "intervalFactor": 2,
+              "legendFormat": "{{name}} {{job}}-{{index}}",
+              "refId": "A",
+              "step": 20
+            }
+          ],
+          "thresholds": [],
+          "timeFrom": null,
+          "timeShift": null,
+          "title": "Top 10 cache misses",
+          "tooltip": {
+            "shared": true,
+            "sort": 0,
+            "value_type": "individual"
+          },
+          "type": "graph",
+          "xaxis": {
+            "buckets": null,
+            "mode": "time",
+            "name": null,
+            "show": true,
+            "values": []
+          },
+          "yaxes": [
+            {
+              "format": "rps",
+              "label": null,
+              "logBase": 1,
+              "max": null,
+              "min": null,
+              "show": true
+            },
+            {
+              "format": "short",
+              "label": null,
+              "logBase": 1,
+              "max": null,
+              "min": null,
+              "show": true
+            }
+          ],
+          "yaxis": {
+            "align": false,
+            "alignLevel": null
+          }
+        },
+        {
+          "aliasColors": {},
+          "bars": false,
+          "dashLength": 10,
+          "dashes": false,
+          "datasource": "${DS_PROMETHEUS}",
+          "fill": 1,
+          "gridPos": {
+            "h": 9,
+            "w": 12,
+            "x": 0,
+            "y": 75
+          },
+          "id": 65,
+          "legend": {
+            "alignAsTable": true,
+            "avg": false,
+            "current": false,
+            "max": false,
+            "min": false,
+            "show": true,
+            "total": false,
+            "values": false
+          },
+          "lines": true,
+          "linewidth": 1,
+          "links": [],
+          "nullPointMode": "null",
+          "percentage": false,
+          "pointradius": 5,
+          "points": false,
+          "renderer": "flot",
+          "seriesOverrides": [],
+          "spaceLength": 10,
+          "stack": false,
+          "steppedLine": false,
+          "targets": [
+            {
+              "expr": "rate(synapse_util_caches_cache:evicted_size{instance=\"$instance\",job=~\"$job\",index=~\"$index\"}[$bucket_size])",
+              "format": "time_series",
+              "intervalFactor": 1,
+              "legendFormat": "{{name}} {{job}}-{{index}}",
+              "refId": "A"
+            }
+          ],
+          "thresholds": [],
+          "timeFrom": null,
+          "timeShift": null,
+          "title": "Cache eviction rate",
+          "tooltip": {
+            "shared": false,
+            "sort": 0,
+            "value_type": "individual"
+          },
+          "transparent": false,
+          "type": "graph",
+          "xaxis": {
+            "buckets": null,
+            "mode": "time",
+            "name": null,
+            "show": true,
+            "values": []
+          },
+          "yaxes": [
+            {
+              "decimals": null,
+              "format": "hertz",
+              "label": "entries / second",
+              "logBase": 1,
+              "max": null,
+              "min": null,
+              "show": true
+            },
+            {
+              "format": "short",
+              "label": null,
+              "logBase": 1,
+              "max": null,
+              "min": null,
+              "show": true
+            }
+          ],
+          "yaxis": {
+            "align": false,
+            "alignLevel": null
+          }
+        }
+      ],
+      "repeat": null,
+      "title": "Caches",
+      "type": "row"
+    },
+    {
+      "collapsed": true,
+      "gridPos": {
+        "h": 1,
+        "w": 24,
+        "x": 0,
+        "y": 26
+      },
+      "id": 62,
+      "panels": [
+        {
+          "aliasColors": {},
+          "bars": false,
+          "dashLength": 10,
+          "dashes": false,
+          "datasource": "${DS_PROMETHEUS}",
+          "fill": 1,
+          "gridPos": {
+            "h": 9,
+            "w": 12,
+            "x": 0,
+            "y": 90
+          },
+          "id": 91,
+          "legend": {
+            "avg": false,
+            "current": false,
+            "max": false,
+            "min": false,
+            "show": true,
+            "total": false,
+            "values": false
+          },
+          "lines": true,
+          "linewidth": 1,
+          "links": [],
+          "nullPointMode": "null",
+          "percentage": false,
+          "pointradius": 5,
+          "points": false,
+          "renderer": "flot",
+          "seriesOverrides": [],
+          "spaceLength": 10,
+          "stack": true,
+          "steppedLine": false,
+          "targets": [
+            {
+              "expr": "rate(python_gc_time_sum{instance=\"$instance\",job=~\"$job\",index=~\"$index\"}[10m])",
+              "format": "time_series",
+              "instant": false,
+              "intervalFactor": 1,
+              "legendFormat": "{{job}}-{{index}} gen {{gen}}",
+              "refId": "A"
+            }
+          ],
+          "thresholds": [],
+          "timeFrom": null,
+          "timeShift": null,
+          "title": "Total GC time by bucket (10m smoothing)",
+          "tooltip": {
+            "shared": true,
+            "sort": 0,
+            "value_type": "individual"
+          },
+          "type": "graph",
+          "xaxis": {
+            "buckets": null,
+            "mode": "time",
+            "name": null,
+            "show": true,
+            "values": []
+          },
+          "yaxes": [
+            {
+              "decimals": null,
+              "format": "percentunit",
+              "label": null,
+              "logBase": 1,
+              "max": null,
+              "min": "0",
+              "show": true
+            },
+            {
+              "format": "short",
+              "label": null,
+              "logBase": 1,
+              "max": null,
+              "min": null,
+              "show": true
+            }
+          ],
+          "yaxis": {
+            "align": false,
+            "alignLevel": null
+          }
+        },
+        {
+          "aliasColors": {},
+          "bars": false,
+          "dashLength": 10,
+          "dashes": false,
+          "datasource": "$datasource",
+          "decimals": 3,
+          "editable": true,
+          "error": false,
+          "fill": 1,
+          "grid": {},
+          "gridPos": {
+            "h": 9,
+            "w": 12,
+            "x": 12,
+            "y": 90
+          },
+          "id": 21,
+          "legend": {
+            "alignAsTable": true,
+            "avg": false,
+            "current": false,
+            "max": false,
+            "min": false,
+            "show": true,
+            "total": false,
+            "values": false
+          },
+          "lines": true,
+          "linewidth": 2,
+          "links": [],
+          "nullPointMode": "null as zero",
+          "percentage": false,
+          "pointradius": 5,
+          "points": false,
+          "renderer": "flot",
+          "seriesOverrides": [],
+          "spaceLength": 10,
+          "stack": false,
+          "steppedLine": false,
+          "targets": [
+            {
+              "expr": "rate(python_gc_time_sum{instance=\"$instance\",job=~\"$job\"}[$bucket_size])/rate(python_gc_time_count[$bucket_size])",
+              "format": "time_series",
+              "intervalFactor": 2,
+              "legendFormat": "{{job}} {{index}} gen {{gen}} ",
+              "refId": "A",
+              "step": 20,
+              "target": ""
+            }
+          ],
+          "thresholds": [],
+          "timeFrom": null,
+          "timeShift": null,
+          "title": "Average GC Time Per Collection",
+          "tooltip": {
+            "shared": false,
+            "sort": 0,
+            "value_type": "cumulative"
+          },
+          "type": "graph",
+          "xaxis": {
+            "buckets": null,
+            "mode": "time",
+            "name": null,
+            "show": true,
+            "values": []
+          },
+          "yaxes": [
+            {
+              "format": "s",
+              "logBase": 1,
+              "max": null,
+              "min": null,
+              "show": true
+            },
+            {
+              "format": "short",
+              "logBase": 1,
+              "max": null,
+              "min": null,
+              "show": true
+            }
+          ],
+          "yaxis": {
+            "align": false,
+            "alignLevel": null
+          }
+        },
+        {
+          "aliasColors": {},
+          "bars": false,
+          "dashLength": 10,
+          "dashes": false,
+          "datasource": "${DS_PROMETHEUS}",
+          "fill": 1,
+          "gridPos": {
+            "h": 9,
+            "w": 12,
+            "x": 0,
+            "y": 99
+          },
+          "id": 89,
+          "legend": {
+            "avg": false,
+            "current": false,
+            "hideEmpty": true,
+            "hideZero": false,
+            "max": false,
+            "min": false,
+            "show": true,
+            "total": false,
+            "values": false
+          },
+          "lines": true,
+          "linewidth": 1,
+          "links": [],
+          "nullPointMode": "null",
+          "percentage": false,
+          "pointradius": 5,
+          "points": false,
+          "renderer": "flot",
+          "seriesOverrides": [],
+          "spaceLength": 10,
+          "stack": false,
+          "steppedLine": false,
+          "targets": [
+            {
+              "expr": "python_gc_counts{job=~\"$job\",index=~\"$index\",instance=\"$instance\"}",
+              "format": "time_series",
+              "intervalFactor": 1,
+              "legendFormat": "{{job}}-{{index}} gen {{gen}}",
+              "refId": "A"
+            }
+          ],
+          "thresholds": [],
+          "timeFrom": null,
+          "timeShift": null,
+          "title": "Currently allocated objects",
+          "tooltip": {
+            "shared": false,
+            "sort": 0,
+            "value_type": "individual"
+          },
+          "type": "graph",
+          "xaxis": {
+            "buckets": null,
+            "mode": "time",
+            "name": null,
+            "show": true,
+            "values": []
+          },
+          "yaxes": [
+            {
+              "format": "short",
+              "label": null,
+              "logBase": 1,
+              "max": null,
+              "min": null,
+              "show": true
+            },
+            {
+              "format": "short",
+              "label": null,
+              "logBase": 1,
+              "max": null,
+              "min": null,
+              "show": true
+            }
+          ],
+          "yaxis": {
+            "align": false,
+            "alignLevel": null
+          }
+        },
+        {
+          "aliasColors": {},
+          "bars": false,
+          "dashLength": 10,
+          "dashes": false,
+          "datasource": "${DS_PROMETHEUS}",
+          "fill": 1,
+          "gridPos": {
+            "h": 9,
+            "w": 12,
+            "x": 12,
+            "y": 99
+          },
+          "id": 93,
+          "legend": {
+            "avg": false,
+            "current": false,
+            "max": false,
+            "min": false,
+            "show": true,
+            "total": false,
+            "values": false
+          },
+          "lines": true,
+          "linewidth": 1,
+          "links": [],
+          "nullPointMode": "null",
+          "percentage": false,
+          "pointradius": 5,
+          "points": false,
+          "renderer": "flot",
+          "seriesOverrides": [],
+          "spaceLength": 10,
+          "stack": false,
+          "steppedLine": false,
+          "targets": [
+            {
+              "expr": "rate(python_gc_unreachable_total{instance=\"$instance\",job=~\"$job\",index=~\"$index\"}[$bucket_size])/rate(python_gc_time_count{instance=\"$instance\",job=~\"$job\",index=~\"$index\"}[$bucket_size])",
+              "format": "time_series",
+              "intervalFactor": 1,
+              "legendFormat": "{{job}}-{{index}} gen {{gen}}",
+              "refId": "A"
+            }
+          ],
+          "thresholds": [],
+          "timeFrom": null,
+          "timeShift": null,
+          "title": "Object counts per collection",
+          "tooltip": {
+            "shared": true,
+            "sort": 0,
+            "value_type": "individual"
+          },
+          "type": "graph",
+          "xaxis": {
+            "buckets": null,
+            "mode": "time",
+            "name": null,
+            "show": true,
+            "values": []
+          },
+          "yaxes": [
+            {
+              "format": "short",
+              "label": null,
+              "logBase": 1,
+              "max": null,
+              "min": null,
+              "show": true
+            },
+            {
+              "format": "short",
+              "label": null,
+              "logBase": 1,
+              "max": null,
+              "min": null,
+              "show": true
+            }
+          ],
+          "yaxis": {
+            "align": false,
+            "alignLevel": null
+          }
+        },
+        {
+          "aliasColors": {},
+          "bars": false,
+          "dashLength": 10,
+          "dashes": false,
+          "datasource": "${DS_PROMETHEUS}",
+          "fill": 1,
+          "gridPos": {
+            "h": 9,
+            "w": 12,
+            "x": 0,
+            "y": 108
+          },
+          "id": 95,
+          "legend": {
+            "avg": false,
+            "current": false,
+            "max": false,
+            "min": false,
+            "show": true,
+            "total": false,
+            "values": false
+          },
+          "lines": true,
+          "linewidth": 1,
+          "links": [],
+          "nullPointMode": "null",
+          "percentage": false,
+          "pointradius": 5,
+          "points": false,
+          "renderer": "flot",
+          "seriesOverrides": [],
+          "spaceLength": 10,
+          "stack": false,
+          "steppedLine": false,
+          "targets": [
+            {
+              "expr": "rate(python_gc_time_count{instance=\"$instance\",job=~\"$job\",index=~\"$index\"}[$bucket_size])",
+              "format": "time_series",
+              "intervalFactor": 1,
+              "legendFormat": "{{job}}-{{index}} gen {{gen}}",
+              "refId": "A"
+            }
+          ],
+          "thresholds": [],
+          "timeFrom": null,
+          "timeShift": null,
+          "title": "GC frequency",
+          "tooltip": {
+            "shared": true,
+            "sort": 0,
+            "value_type": "individual"
+          },
+          "type": "graph",
+          "xaxis": {
+            "buckets": null,
+            "mode": "time",
+            "name": null,
+            "show": true,
+            "values": []
+          },
+          "yaxes": [
+            {
+              "format": "hertz",
+              "label": null,
+              "logBase": 1,
+              "max": null,
+              "min": null,
+              "show": true
+            },
+            {
+              "format": "short",
+              "label": null,
+              "logBase": 1,
+              "max": null,
+              "min": null,
+              "show": true
+            }
+          ],
+          "yaxis": {
+            "align": false,
+            "alignLevel": null
+          }
+        },
+        {
+          "cards": {
+            "cardPadding": 0,
+            "cardRound": null
+          },
+          "color": {
+            "cardColor": "#b4ff00",
+            "colorScale": "sqrt",
+            "colorScheme": "interpolateSpectral",
+            "exponent": 0.5,
+            "max": null,
+            "min": 0,
+            "mode": "spectrum"
+          },
+          "dataFormat": "tsbuckets",
+          "datasource": "${DS_PROMETHEUS}",
+          "gridPos": {
+            "h": 9,
+            "w": 12,
+            "x": 12,
+            "y": 108
+          },
+          "heatmap": {},
+          "highlightCards": true,
+          "id": 87,
+          "legend": {
+            "show": true
+          },
+          "links": [],
+          "targets": [
+            {
+              "expr": "sum(rate(python_gc_time_bucket[$bucket_size])) by (le)",
+              "format": "heatmap",
+              "intervalFactor": 1,
+              "legendFormat": "{{le}}",
+              "refId": "A"
+            }
+          ],
+          "title": "GC durations",
+          "tooltip": {
+            "show": true,
+            "showHistogram": false
+          },
+          "type": "heatmap",
+          "xAxis": {
+            "show": true
+          },
+          "xBucketNumber": null,
+          "xBucketSize": null,
+          "yAxis": {
+            "decimals": null,
+            "format": "s",
+            "logBase": 1,
+            "max": null,
+            "min": null,
+            "show": true,
+            "splitFactor": null
+          },
+          "yBucketBound": "auto",
+          "yBucketNumber": null,
+          "yBucketSize": null
+        }
+      ],
+      "repeat": null,
+      "title": "GC",
+      "type": "row"
+    },
+    {
+      "collapsed": true,
+      "gridPos": {
+        "h": 1,
+        "w": 24,
+        "x": 0,
+        "y": 27
+      },
+      "id": 63,
+      "panels": [
+        {
+          "aliasColors": {},
+          "bars": false,
+          "dashLength": 10,
+          "dashes": false,
+          "datasource": "${DS_PROMETHEUS}",
+          "fill": 1,
+          "gridPos": {
+            "h": 7,
+            "w": 12,
+            "x": 0,
+            "y": 19
+          },
+          "id": 2,
+          "legend": {
+            "avg": false,
+            "current": false,
+            "max": false,
+            "min": false,
+            "show": true,
+            "total": false,
+            "values": false
+          },
+          "lines": true,
+          "linewidth": 1,
+          "links": [],
+          "nullPointMode": "null",
+          "percentage": false,
+          "pointradius": 5,
+          "points": false,
+          "renderer": "flot",
+          "seriesOverrides": [],
+          "spaceLength": 10,
+          "stack": false,
+          "steppedLine": false,
+          "targets": [
+            {
+              "expr": "rate(synapse_replication_tcp_resource_user_sync{instance=\"$instance\",job=~\"$job\",index=~\"$index\"}[$bucket_size])",
+              "format": "time_series",
+              "intervalFactor": 2,
+              "legendFormat": "user started/stopped syncing",
+              "refId": "A",
+              "step": 20
+            },
+            {
+              "expr": "rate(synapse_replication_tcp_resource_federation_ack{instance=\"$instance\",job=~\"$job\",index=~\"$index\"}[$bucket_size])",
+              "format": "time_series",
+              "intervalFactor": 2,
+              "legendFormat": "federation ack",
+              "refId": "B",
+              "step": 20
+            },
+            {
+              "expr": "rate(synapse_replication_tcp_resource_remove_pusher{instance=\"$instance\",job=~\"$job\",index=~\"$index\"}[$bucket_size])",
+              "format": "time_series",
+              "intervalFactor": 2,
+              "legendFormat": "remove pusher",
+              "refId": "C",
+              "step": 20
+            },
+            {
+              "expr": "rate(synapse_replication_tcp_resource_invalidate_cache{instance=\"$instance\",job=~\"$job\",index=~\"$index\"}[$bucket_size])",
+              "format": "time_series",
+              "intervalFactor": 2,
+              "legendFormat": "invalidate cache",
+              "refId": "D",
+              "step": 20
+            },
+            {
+              "expr": "rate(synapse_replication_tcp_resource_user_ip_cache{instance=\"$instance\",job=~\"$job\",index=~\"$index\"}[$bucket_size])",
+              "format": "time_series",
+              "intervalFactor": 2,
+              "legendFormat": "user ip cache",
+              "refId": "E",
+              "step": 20
+            }
+          ],
+          "thresholds": [],
+          "timeFrom": null,
+          "timeShift": null,
+          "title": "Rate of events on replication master",
+          "tooltip": {
+            "shared": true,
+            "sort": 0,
+            "value_type": "individual"
+          },
+          "type": "graph",
+          "xaxis": {
+            "buckets": null,
+            "mode": "time",
+            "name": null,
+            "show": true,
+            "values": []
+          },
+          "yaxes": [
+            {
+              "format": "hertz",
+              "label": null,
+              "logBase": 1,
+              "max": null,
+              "min": null,
+              "show": true
+            },
+            {
+              "format": "short",
+              "label": null,
+              "logBase": 1,
+              "max": null,
+              "min": null,
+              "show": true
+            }
+          ]
+        },
+        {
+          "aliasColors": {},
+          "bars": false,
+          "dashLength": 10,
+          "dashes": false,
+          "datasource": "${DS_PROMETHEUS}",
+          "fill": 1,
+          "gridPos": {
+            "h": 7,
+            "w": 12,
+            "x": 12,
+            "y": 19
+          },
+          "id": 41,
+          "legend": {
+            "avg": false,
+            "current": false,
+            "max": false,
+            "min": false,
+            "show": true,
+            "total": false,
+            "values": false
+          },
+          "lines": true,
+          "linewidth": 1,
+          "links": [],
+          "nullPointMode": "null",
+          "percentage": false,
+          "pointradius": 5,
+          "points": false,
+          "renderer": "flot",
+          "seriesOverrides": [],
+          "spaceLength": 10,
+          "stack": false,
+          "steppedLine": false,
+          "targets": [
+            {
+              "expr": "rate(synapse_replication_tcp_resource_stream_updates{job=~\"$job\",index=~\"$index\",instance=\"$instance\"}[$bucket_size])",
+              "format": "time_series",
+              "interval": "",
+              "intervalFactor": 2,
+              "legendFormat": "{{stream_name}}",
+              "refId": "A",
+              "step": 20
+            }
+          ],
+          "thresholds": [],
+          "timeFrom": null,
+          "timeShift": null,
+          "title": "Outgoing stream updates",
+          "tooltip": {
+            "shared": true,
+            "sort": 0,
+            "value_type": "individual"
+          },
+          "type": "graph",
+          "xaxis": {
+            "buckets": null,
+            "mode": "time",
+            "name": null,
+            "show": true,
+            "values": []
+          },
+          "yaxes": [
+            {
+              "format": "hertz",
+              "label": null,
+              "logBase": 1,
+              "max": null,
+              "min": null,
+              "show": true
+            },
+            {
+              "format": "short",
+              "label": null,
+              "logBase": 1,
+              "max": null,
+              "min": null,
+              "show": true
+            }
+          ]
+        },
+        {
+          "aliasColors": {},
+          "bars": false,
+          "dashLength": 10,
+          "dashes": false,
+          "datasource": "${DS_PROMETHEUS}",
+          "fill": 1,
+          "gridPos": {
+            "h": 7,
+            "w": 12,
+            "x": 0,
+            "y": 26
+          },
+          "id": 42,
+          "legend": {
+            "avg": false,
+            "current": false,
+            "max": false,
+            "min": false,
+            "show": true,
+            "total": false,
+            "values": false
+          },
+          "lines": true,
+          "linewidth": 1,
+          "links": [],
+          "nullPointMode": "null",
+          "percentage": false,
+          "pointradius": 5,
+          "points": false,
+          "renderer": "flot",
+          "seriesOverrides": [],
+          "spaceLength": 10,
+          "stack": false,
+          "steppedLine": false,
+          "targets": [
+            {
+              "expr": "sum (rate(synapse_replication_tcp_protocol_inbound_commands{instance=\"$instance\",job=~\"$job\",index=~\"$index\"}[$bucket_size])) without (name, conn_id)",
+              "format": "time_series",
+              "intervalFactor": 2,
+              "legendFormat": "{{job}}-{{index}} {{command}}",
+              "refId": "A",
+              "step": 20
+            }
+          ],
+          "thresholds": [],
+          "timeFrom": null,
+          "timeShift": null,
+          "title": "Rate of incoming commands",
+          "tooltip": {
+            "shared": true,
+            "sort": 0,
+            "value_type": "individual"
+          },
+          "type": "graph",
+          "xaxis": {
+            "buckets": null,
+            "mode": "time",
+            "name": null,
+            "show": true,
+            "values": []
+          },
+          "yaxes": [
+            {
+              "format": "hertz",
+              "label": null,
+              "logBase": 1,
+              "max": null,
+              "min": null,
+              "show": true
+            },
+            {
+              "format": "short",
+              "label": null,
+              "logBase": 1,
+              "max": null,
+              "min": null,
+              "show": true
+            }
+          ]
+        },
+        {
+          "aliasColors": {},
+          "bars": false,
+          "dashLength": 10,
+          "dashes": false,
+          "datasource": "${DS_PROMETHEUS}",
+          "fill": 1,
+          "gridPos": {
+            "h": 7,
+            "w": 12,
+            "x": 12,
+            "y": 26
+          },
+          "id": 43,
+          "legend": {
+            "avg": false,
+            "current": false,
+            "max": false,
+            "min": false,
+            "show": true,
+            "total": false,
+            "values": false
+          },
+          "lines": true,
+          "linewidth": 1,
+          "links": [],
+          "nullPointMode": "null",
+          "percentage": false,
+          "pointradius": 5,
+          "points": false,
+          "renderer": "flot",
+          "seriesOverrides": [],
+          "spaceLength": 10,
+          "stack": false,
+          "steppedLine": false,
+          "targets": [
+            {
+              "expr": "sum (rate(synapse_replication_tcp_protocol_outbound_commands{instance=\"$instance\",job=~\"$job\",index=~\"$index\"}[$bucket_size])) without (name, conn_id)",
+              "format": "time_series",
+              "intervalFactor": 2,
+              "legendFormat": "{{job}}-{{index}} {{command}}",
+              "refId": "A",
+              "step": 20
+            }
+          ],
+          "thresholds": [],
+          "timeFrom": null,
+          "timeShift": null,
+          "title": "Rate of outgoing commands",
+          "tooltip": {
+            "shared": true,
+            "sort": 0,
+            "value_type": "individual"
+          },
+          "type": "graph",
+          "xaxis": {
+            "buckets": null,
+            "mode": "time",
+            "name": null,
+            "show": true,
+            "values": []
+          },
+          "yaxes": [
+            {
+              "format": "hertz",
+              "label": null,
+              "logBase": 1,
+              "max": null,
+              "min": null,
+              "show": true
+            },
+            {
+              "format": "short",
+              "label": null,
+              "logBase": 1,
+              "max": null,
+              "min": null,
+              "show": true
+            }
+          ]
+        }
+      ],
+      "repeat": null,
+      "title": "Replication",
+      "type": "row"
+    },
+    {
+      "collapsed": true,
+      "gridPos": {
+        "h": 1,
+        "w": 24,
+        "x": 0,
+        "y": 28
+      },
+      "id": 69,
+      "panels": [
+        {
+          "aliasColors": {},
+          "bars": false,
+          "dashLength": 10,
+          "dashes": false,
+          "datasource": "${DS_PROMETHEUS}",
+          "fill": 1,
+          "gridPos": {
+            "h": 9,
+            "w": 12,
+            "x": 0,
+            "y": 11
+          },
+          "id": 67,
+          "legend": {
+            "avg": false,
+            "current": false,
+            "max": false,
+            "min": false,
+            "show": true,
+            "total": false,
+            "values": false
+          },
+          "lines": true,
+          "linewidth": 1,
+          "links": [],
+          "nullPointMode": "null",
+          "percentage": false,
+          "pointradius": 5,
+          "points": false,
+          "renderer": "flot",
+          "seriesOverrides": [],
+          "spaceLength": 10,
+          "stack": false,
+          "steppedLine": false,
+          "targets": [
+            {
+              "expr": " synapse_event_persisted_position{instance=\"$instance\"} - ignoring(index, job, name) group_right(instance) synapse_event_processing_positions{instance=\"$instance\",job=~\"$job\",index=~\"$index\"}",
+              "format": "time_series",
+              "interval": "",
+              "intervalFactor": 1,
+              "legendFormat": "{{job}}-{{index}}",
+              "refId": "A"
+            }
+          ],
+          "thresholds": [],
+          "timeFrom": null,
+          "timeShift": null,
+          "title": "Event processing lag",
+          "tooltip": {
+            "shared": true,
+            "sort": 0,
+            "value_type": "individual"
+          },
+          "type": "graph",
+          "xaxis": {
+            "buckets": null,
+            "mode": "time",
+            "name": null,
+            "show": true,
+            "values": []
+          },
+          "yaxes": [
+            {
+              "format": "short",
+              "label": "events",
+              "logBase": 1,
+              "max": null,
+              "min": null,
+              "show": true
+            },
+            {
+              "format": "short",
+              "label": null,
+              "logBase": 1,
+              "max": null,
+              "min": null,
+              "show": true
+            }
+          ]
+        },
+        {
+          "aliasColors": {},
+          "bars": false,
+          "dashLength": 10,
+          "dashes": false,
+          "datasource": "${DS_PROMETHEUS}",
+          "fill": 1,
+          "gridPos": {
+            "h": 9,
+            "w": 12,
+            "x": 12,
+            "y": 11
+          },
+          "id": 71,
+          "legend": {
+            "avg": false,
+            "current": false,
+            "max": false,
+            "min": false,
+            "show": true,
+            "total": false,
+            "values": false
+          },
+          "lines": true,
+          "linewidth": 1,
+          "links": [],
+          "nullPointMode": "null",
+          "percentage": false,
+          "pointradius": 5,
+          "points": false,
+          "renderer": "flot",
+          "seriesOverrides": [],
+          "spaceLength": 10,
+          "stack": false,
+          "steppedLine": false,
+          "targets": [
+            {
+              "expr": "time()*1000-synapse_event_processing_last_ts{instance=\"$instance\",job=~\"$job\",index=~\"$index\"}",
+              "format": "time_series",
+              "hide": false,
+              "intervalFactor": 1,
+              "legendFormat": "{{job}}-{{index}} {{name}}",
+              "refId": "B"
+            }
+          ],
+          "thresholds": [],
+          "timeFrom": null,
+          "timeShift": null,
+          "title": "Age of last processed event",
+          "tooltip": {
+            "shared": true,
+            "sort": 0,
+            "value_type": "individual"
+          },
+          "type": "graph",
+          "xaxis": {
+            "buckets": null,
+            "mode": "time",
+            "name": null,
+            "show": true,
+            "values": []
+          },
+          "yaxes": [
+            {
+              "format": "ms",
+              "label": null,
+              "logBase": 1,
+              "max": null,
+              "min": null,
+              "show": true
+            },
+            {
+              "format": "short",
+              "label": null,
+              "logBase": 1,
+              "max": null,
+              "min": null,
+              "show": true
+            }
+          ]
+        }
+      ],
+      "title": "Event processing loop positions",
+      "type": "row"
+    }
+  ],
+  "refresh": "1m",
+  "schemaVersion": 16,
+  "style": "dark",
+  "tags": [
+    "matrix"
+  ],
+  "templating": {
+    "list": [
+      {
+        "current": {
+          "text": "Prometheus",
+          "value": "Prometheus"
+        },
+        "hide": 0,
+        "label": null,
+        "name": "datasource",
+        "options": [],
+        "query": "prometheus",
+        "refresh": 1,
+        "regex": "",
+        "type": "datasource"
+      },
+      {
+        "allFormat": "glob",
+        "auto": true,
+        "auto_count": 100,
+        "auto_min": "30s",
+        "current": {
+          "text": "auto",
+          "value": "$__auto_interval_bucket_size"
+        },
+        "datasource": null,
+        "hide": 0,
+        "includeAll": false,
+        "label": "Bucket Size",
+        "multi": false,
+        "multiFormat": "glob",
+        "name": "bucket_size",
+        "options": [
+          {
+            "selected": true,
+            "text": "auto",
+            "value": "$__auto_interval_bucket_size"
+          },
+          {
+            "selected": false,
+            "text": "30s",
+            "value": "30s"
+          },
+          {
+            "selected": false,
+            "text": "1m",
+            "value": "1m"
+          },
+          {
+            "selected": false,
+            "text": "2m",
+            "value": "2m"
+          },
+          {
+            "selected": false,
+            "text": "5m",
+            "value": "5m"
+          }
+        ],
+        "query": "30s,1m,2m,5m",
+        "refresh": 2,
+        "type": "interval"
+      },
+      {
+        "allValue": null,
+        "current": {},
+        "datasource": "$datasource",
+        "hide": 0,
+        "includeAll": false,
+        "label": null,
+        "multi": false,
+        "name": "instance",
+        "options": [],
+        "query": "label_values(process_cpu_user_seconds_total{job=~\"synapse.*\"}, instance)",
+        "refresh": 2,
+        "regex": "",
+        "sort": 0,
+        "tagValuesQuery": "",
+        "tags": [],
+        "tagsQuery": "",
+        "type": "query",
+        "useTags": false
+      },
+      {
+        "allFormat": "regex wildcard",
+        "allValue": "",
+        "current": {},
+        "datasource": "$datasource",
+        "hide": 0,
+        "hideLabel": false,
+        "includeAll": true,
+        "label": "Job",
+        "multi": true,
+        "multiFormat": "regex values",
+        "name": "job",
+        "options": [],
+        "query": "label_values(process_cpu_user_seconds_total{job=~\"synapse.*\"}, job)",
+        "refresh": 2,
+        "refresh_on_load": false,
+        "regex": "",
+        "sort": 1,
+        "tagValuesQuery": "",
+        "tags": [],
+        "tagsQuery": "",
+        "type": "query",
+        "useTags": false
+      },
+      {
+        "allFormat": "regex wildcard",
+        "allValue": ".*",
+        "current": {},
+        "datasource": "$datasource",
+        "hide": 0,
+        "hideLabel": false,
+        "includeAll": true,
+        "label": "",
+        "multi": true,
+        "multiFormat": "regex values",
+        "name": "index",
+        "options": [],
+        "query": "label_values(process_cpu_user_seconds_total{job=~\"synapse.*\"}, index)",
+        "refresh": 2,
+        "refresh_on_load": false,
+        "regex": "",
+        "sort": 3,
+        "tagValuesQuery": "",
+        "tags": [],
+        "tagsQuery": "",
+        "type": "query",
+        "useTags": false
+      }
+    ]
+  },
+  "time": {
+    "from": "now-1h",
+    "to": "now"
+  },
+  "timepicker": {
+    "now": true,
+    "refresh_intervals": [
+      "5s",
+      "10s",
+      "30s",
+      "1m",
+      "5m",
+      "15m",
+      "30m",
+      "1h",
+      "2h",
+      "1d"
+    ],
+    "time_options": [
+      "5m",
+      "15m",
+      "1h",
+      "6h",
+      "12h",
+      "24h",
+      "2d",
+      "7d",
+      "30d"
+    ]
+  },
+  "timezone": "",
+  "title": "Synapse",
+  "uid": "000000012",
+  "version": 125
+}
\ No newline at end of file
diff --git a/synapse/federation/federation_server.py b/synapse/federation/federation_server.py
index e501251b6e..657935d1ac 100644
--- a/synapse/federation/federation_server.py
+++ b/synapse/federation/federation_server.py
@@ -207,10 +207,6 @@ class FederationServer(FederationBase):
                     edu.content
                 )
 
-        pdu_failures = getattr(transaction, "pdu_failures", [])
-        for fail in pdu_failures:
-            logger.info("Got failure %r", fail)
-
         response = {
             "pdus": pdu_results,
         }
diff --git a/synapse/federation/send_queue.py b/synapse/federation/send_queue.py
index 5157c3860d..0bb468385d 100644
--- a/synapse/federation/send_queue.py
+++ b/synapse/federation/send_queue.py
@@ -62,8 +62,6 @@ class FederationRemoteSendQueue(object):
 
         self.edus = SortedDict()  # stream position -> Edu
 
-        self.failures = SortedDict()  # stream position -> (destination, Failure)
-
         self.device_messages = SortedDict()  # stream position -> destination
 
         self.pos = 1
@@ -79,7 +77,7 @@ class FederationRemoteSendQueue(object):
 
         for queue_name in [
             "presence_map", "presence_changed", "keyed_edu", "keyed_edu_changed",
-            "edus", "failures", "device_messages", "pos_time",
+            "edus", "device_messages", "pos_time",
         ]:
             register(queue_name, getattr(self, queue_name))
 
@@ -149,12 +147,6 @@ class FederationRemoteSendQueue(object):
             for key in keys[:i]:
                 del self.edus[key]
 
-            # Delete things out of failure map
-            keys = self.failures.keys()
-            i = self.failures.bisect_left(position_to_delete)
-            for key in keys[:i]:
-                del self.failures[key]
-
             # Delete things out of device map
             keys = self.device_messages.keys()
             i = self.device_messages.bisect_left(position_to_delete)
@@ -204,13 +196,6 @@ class FederationRemoteSendQueue(object):
 
         self.notifier.on_new_replication_data()
 
-    def send_failure(self, failure, destination):
-        """As per TransactionQueue"""
-        pos = self._next_pos()
-
-        self.failures[pos] = (destination, str(failure))
-        self.notifier.on_new_replication_data()
-
     def send_device_messages(self, destination):
         """As per TransactionQueue"""
         pos = self._next_pos()
@@ -285,17 +270,6 @@ class FederationRemoteSendQueue(object):
         for (pos, edu) in edus:
             rows.append((pos, EduRow(edu)))
 
-        # Fetch changed failures
-        i = self.failures.bisect_right(from_token)
-        j = self.failures.bisect_right(to_token) + 1
-        failures = self.failures.items()[i:j]
-
-        for (pos, (destination, failure)) in failures:
-            rows.append((pos, FailureRow(
-                destination=destination,
-                failure=failure,
-            )))
-
         # Fetch changed device messages
         i = self.device_messages.bisect_right(from_token)
         j = self.device_messages.bisect_right(to_token) + 1
@@ -417,34 +391,6 @@ class EduRow(BaseFederationRow, namedtuple("EduRow", (
         buff.edus.setdefault(self.edu.destination, []).append(self.edu)
 
 
-class FailureRow(BaseFederationRow, namedtuple("FailureRow", (
-    "destination",  # str
-    "failure",
-))):
-    """Streams failures to a remote server. Failures are issued when there was
-    something wrong with a transaction the remote sent us, e.g. it included
-    an event that was invalid.
-    """
-
-    TypeId = "f"
-
-    @staticmethod
-    def from_data(data):
-        return FailureRow(
-            destination=data["destination"],
-            failure=data["failure"],
-        )
-
-    def to_data(self):
-        return {
-            "destination": self.destination,
-            "failure": self.failure,
-        }
-
-    def add_to_buffer(self, buff):
-        buff.failures.setdefault(self.destination, []).append(self.failure)
-
-
 class DeviceRow(BaseFederationRow, namedtuple("DeviceRow", (
     "destination",  # str
 ))):
@@ -471,7 +417,6 @@ TypeToRow = {
         PresenceRow,
         KeyedEduRow,
         EduRow,
-        FailureRow,
         DeviceRow,
     )
 }
@@ -481,7 +426,6 @@ ParsedFederationStreamData = namedtuple("ParsedFederationStreamData", (
     "presence",  # list(UserPresenceState)
     "keyed_edus",  # dict of destination -> { key -> Edu }
     "edus",  # dict of destination -> [Edu]
-    "failures",  # dict of destination -> [failures]
     "device_destinations",  # set of destinations
 ))
 
@@ -503,7 +447,6 @@ def process_rows_for_federation(transaction_queue, rows):
         presence=[],
         keyed_edus={},
         edus={},
-        failures={},
         device_destinations=set(),
     )
 
@@ -532,9 +475,5 @@ def process_rows_for_federation(transaction_queue, rows):
                 edu.destination, edu.edu_type, edu.content, key=None,
             )
 
-    for destination, failure_list in iteritems(buff.failures):
-        for failure in failure_list:
-            transaction_queue.send_failure(destination, failure)
-
     for destination in buff.device_destinations:
         transaction_queue.send_device_messages(destination)
diff --git a/synapse/federation/transaction_queue.py b/synapse/federation/transaction_queue.py
index 6996d6b695..78f9d40a3a 100644
--- a/synapse/federation/transaction_queue.py
+++ b/synapse/federation/transaction_queue.py
@@ -116,9 +116,6 @@ class TransactionQueue(object):
             ),
         )
 
-        # destination -> list of tuple(failure, deferred)
-        self.pending_failures_by_dest = {}
-
         # destination -> stream_id of last successfully sent to-device message.
         # NB: may be a long or an int.
         self.last_device_stream_id_by_dest = {}
@@ -382,19 +379,6 @@ class TransactionQueue(object):
 
         self._attempt_new_transaction(destination)
 
-    def send_failure(self, failure, destination):
-        if destination == self.server_name or destination == "localhost":
-            return
-
-        if not self.can_send_to(destination):
-            return
-
-        self.pending_failures_by_dest.setdefault(
-            destination, []
-        ).append(failure)
-
-        self._attempt_new_transaction(destination)
-
     def send_device_messages(self, destination):
         if destination == self.server_name or destination == "localhost":
             return
@@ -469,7 +453,6 @@ class TransactionQueue(object):
                 pending_pdus = self.pending_pdus_by_dest.pop(destination, [])
                 pending_edus = self.pending_edus_by_dest.pop(destination, [])
                 pending_presence = self.pending_presence_by_dest.pop(destination, {})
-                pending_failures = self.pending_failures_by_dest.pop(destination, [])
 
                 pending_edus.extend(
                     self.pending_edus_keyed_by_dest.pop(destination, {}).values()
@@ -497,7 +480,7 @@ class TransactionQueue(object):
                     logger.debug("TX [%s] len(pending_pdus_by_dest[dest]) = %d",
                                  destination, len(pending_pdus))
 
-                if not pending_pdus and not pending_edus and not pending_failures:
+                if not pending_pdus and not pending_edus:
                     logger.debug("TX [%s] Nothing to send", destination)
                     self.last_device_stream_id_by_dest[destination] = (
                         device_stream_id
@@ -507,7 +490,7 @@ class TransactionQueue(object):
                 # END CRITICAL SECTION
 
                 success = yield self._send_new_transaction(
-                    destination, pending_pdus, pending_edus, pending_failures,
+                    destination, pending_pdus, pending_edus,
                 )
                 if success:
                     sent_transactions_counter.inc()
@@ -584,14 +567,12 @@ class TransactionQueue(object):
 
     @measure_func("_send_new_transaction")
     @defer.inlineCallbacks
-    def _send_new_transaction(self, destination, pending_pdus, pending_edus,
-                              pending_failures):
+    def _send_new_transaction(self, destination, pending_pdus, pending_edus):
 
         # Sort based on the order field
         pending_pdus.sort(key=lambda t: t[1])
         pdus = [x[0] for x in pending_pdus]
         edus = pending_edus
-        failures = [x.get_dict() for x in pending_failures]
 
         success = True
 
@@ -601,11 +582,10 @@ class TransactionQueue(object):
 
         logger.debug(
             "TX [%s] {%s} Attempting new transaction"
-            " (pdus: %d, edus: %d, failures: %d)",
+            " (pdus: %d, edus: %d)",
             destination, txn_id,
             len(pdus),
             len(edus),
-            len(failures)
         )
 
         logger.debug("TX [%s] Persisting transaction...", destination)
@@ -617,7 +597,6 @@ class TransactionQueue(object):
             destination=destination,
             pdus=pdus,
             edus=edus,
-            pdu_failures=failures,
         )
 
         self._next_txn_id += 1
@@ -627,12 +606,11 @@ class TransactionQueue(object):
         logger.debug("TX [%s] Persisted transaction", destination)
         logger.info(
             "TX [%s] {%s} Sending transaction [%s],"
-            " (PDUs: %d, EDUs: %d, failures: %d)",
+            " (PDUs: %d, EDUs: %d)",
             destination, txn_id,
             transaction.transaction_id,
             len(pdus),
             len(edus),
-            len(failures),
         )
 
         # Actually send the transaction
diff --git a/synapse/federation/transport/server.py b/synapse/federation/transport/server.py
index 8574898f0c..3b5ea9515a 100644
--- a/synapse/federation/transport/server.py
+++ b/synapse/federation/transport/server.py
@@ -283,11 +283,10 @@ class FederationSendServlet(BaseFederationServlet):
             )
 
             logger.info(
-                "Received txn %s from %s. (PDUs: %d, EDUs: %d, failures: %d)",
+                "Received txn %s from %s. (PDUs: %d, EDUs: %d)",
                 transaction_id, origin,
                 len(transaction_data.get("pdus", [])),
                 len(transaction_data.get("edus", [])),
-                len(transaction_data.get("failures", [])),
             )
 
             # We should ideally be getting this from the security layer.
diff --git a/synapse/federation/units.py b/synapse/federation/units.py
index bb1b3b13f7..c5ab14314e 100644
--- a/synapse/federation/units.py
+++ b/synapse/federation/units.py
@@ -73,7 +73,6 @@ class Transaction(JsonEncodedObject):
         "previous_ids",
         "pdus",
         "edus",
-        "pdu_failures",
     ]
 
     internal_keys = [
diff --git a/synapse/handlers/federation.py b/synapse/handlers/federation.py
index 49068c06d9..91d8def08b 100644
--- a/synapse/handlers/federation.py
+++ b/synapse/handlers/federation.py
@@ -76,7 +76,7 @@ class FederationHandler(BaseHandler):
         self.hs = hs
 
         self.store = hs.get_datastore()
-        self.replication_layer = hs.get_federation_client()
+        self.federation_client = hs.get_federation_client()
         self.state_handler = hs.get_state_handler()
         self.server_name = hs.hostname
         self.keyring = hs.get_keyring()
@@ -255,7 +255,7 @@ class FederationHandler(BaseHandler):
                     # know about
                     for p in prevs - seen:
                         state, got_auth_chain = (
-                            yield self.replication_layer.get_state_for_room(
+                            yield self.federation_client.get_state_for_room(
                                 origin, pdu.room_id, p
                             )
                         )
@@ -338,7 +338,7 @@ class FederationHandler(BaseHandler):
         #
         # see https://github.com/matrix-org/synapse/pull/1744
 
-        missing_events = yield self.replication_layer.get_missing_events(
+        missing_events = yield self.federation_client.get_missing_events(
             origin,
             pdu.room_id,
             earliest_events_ids=list(latest),
@@ -522,7 +522,7 @@ class FederationHandler(BaseHandler):
         if dest == self.server_name:
             raise SynapseError(400, "Can't backfill from self.")
 
-        events = yield self.replication_layer.backfill(
+        events = yield self.federation_client.backfill(
             dest,
             room_id,
             limit=limit,
@@ -570,7 +570,7 @@ class FederationHandler(BaseHandler):
         state_events = {}
         events_to_state = {}
         for e_id in edges:
-            state, auth = yield self.replication_layer.get_state_for_room(
+            state, auth = yield self.federation_client.get_state_for_room(
                 destination=dest,
                 room_id=room_id,
                 event_id=e_id
@@ -612,7 +612,7 @@ class FederationHandler(BaseHandler):
                 results = yield logcontext.make_deferred_yieldable(defer.gatherResults(
                     [
                         logcontext.run_in_background(
-                            self.replication_layer.get_pdu,
+                            self.federation_client.get_pdu,
                             [dest],
                             event_id,
                             outlier=True,
@@ -893,7 +893,7 @@ class FederationHandler(BaseHandler):
 
         Invites must be signed by the invitee's server before distribution.
         """
-        pdu = yield self.replication_layer.send_invite(
+        pdu = yield self.federation_client.send_invite(
             destination=target_host,
             room_id=event.room_id,
             event_id=event.event_id,
@@ -955,7 +955,7 @@ class FederationHandler(BaseHandler):
                 target_hosts.insert(0, origin)
             except ValueError:
                 pass
-            ret = yield self.replication_layer.send_join(target_hosts, event)
+            ret = yield self.federation_client.send_join(target_hosts, event)
 
             origin = ret["origin"]
             state = ret["state"]
@@ -1211,7 +1211,7 @@ class FederationHandler(BaseHandler):
         except ValueError:
             pass
 
-        yield self.replication_layer.send_leave(
+        yield self.federation_client.send_leave(
             target_hosts,
             event
         )
@@ -1234,7 +1234,7 @@ class FederationHandler(BaseHandler):
     @defer.inlineCallbacks
     def _make_and_verify_event(self, target_hosts, room_id, user_id, membership,
                                content={},):
-        origin, pdu = yield self.replication_layer.make_membership_event(
+        origin, pdu = yield self.federation_client.make_membership_event(
             target_hosts,
             room_id,
             user_id,
@@ -1567,7 +1567,7 @@ class FederationHandler(BaseHandler):
                     missing_auth_events.add(e_id)
 
         for e_id in missing_auth_events:
-            m_ev = yield self.replication_layer.get_pdu(
+            m_ev = yield self.federation_client.get_pdu(
                 [origin],
                 e_id,
                 outlier=True,
@@ -1777,7 +1777,7 @@ class FederationHandler(BaseHandler):
             logger.info("Missing auth: %s", missing_auth)
             # If we don't have all the auth events, we need to get them.
             try:
-                remote_auth_chain = yield self.replication_layer.get_event_auth(
+                remote_auth_chain = yield self.federation_client.get_event_auth(
                     origin, event.room_id, event.event_id
                 )
 
@@ -1893,7 +1893,7 @@ class FederationHandler(BaseHandler):
 
                 try:
                     # 2. Get remote difference.
-                    result = yield self.replication_layer.query_auth(
+                    result = yield self.federation_client.query_auth(
                         origin,
                         event.room_id,
                         event.event_id,
@@ -2192,7 +2192,7 @@ class FederationHandler(BaseHandler):
             yield member_handler.send_membership_event(None, event, context)
         else:
             destinations = set(x.split(":", 1)[-1] for x in (sender_user_id, room_id))
-            yield self.replication_layer.forward_third_party_invite(
+            yield self.federation_client.forward_third_party_invite(
                 destinations,
                 room_id,
                 event_dict,
diff --git a/synapse/secrets.py b/synapse/secrets.py
index f397daaa5e..f05e9ea535 100644
--- a/synapse/secrets.py
+++ b/synapse/secrets.py
@@ -20,17 +20,16 @@ See https://docs.python.org/3/library/secrets.html#module-secrets for the API
 used in Python 3.6, and the API emulated in Python 2.7.
 """
 
-import six
+import sys
 
-if six.PY3:
+# secrets is available since python 3.6
+if sys.version_info[0:2] >= (3, 6):
     import secrets
 
     def Secrets():
         return secrets
 
-
 else:
-
     import os
     import binascii
 
diff --git a/synapse/storage/__init__.py b/synapse/storage/__init__.py
index 3ce1b34aea..12282a8b3d 100644
--- a/synapse/storage/__init__.py
+++ b/synapse/storage/__init__.py
@@ -66,6 +66,7 @@ class DataStore(RoomMemberStore, RoomStore,
                 PresenceStore, TransactionStore,
                 DirectoryStore, KeyStore, StateStore, SignatureStore,
                 ApplicationServiceStore,
+                EventsStore,
                 EventFederationStore,
                 MediaRepositoryStore,
                 RejectionsStore,
@@ -73,7 +74,6 @@ class DataStore(RoomMemberStore, RoomStore,
                 PusherStore,
                 PushRuleStore,
                 ApplicationServiceTransactionStore,
-                EventsStore,
                 ReceiptsStore,
                 EndToEndKeyStore,
                 SearchStore,
diff --git a/synapse/storage/appservice.py b/synapse/storage/appservice.py
index 9f12b360bc..31248d5e06 100644
--- a/synapse/storage/appservice.py
+++ b/synapse/storage/appservice.py
@@ -22,7 +22,7 @@ from twisted.internet import defer
 
 from synapse.appservice import AppServiceTransaction
 from synapse.config.appservice import load_appservices
-from synapse.storage.events import EventsWorkerStore
+from synapse.storage.events_worker import EventsWorkerStore
 
 from ._base import SQLBaseStore
 
diff --git a/synapse/storage/event_federation.py b/synapse/storage/event_federation.py
index 5d3ee90017..8bd35df119 100644
--- a/synapse/storage/event_federation.py
+++ b/synapse/storage/event_federation.py
@@ -25,7 +25,7 @@ from twisted.internet import defer
 from synapse.api.errors import StoreError
 from synapse.metrics.background_process_metrics import run_as_background_process
 from synapse.storage._base import SQLBaseStore
-from synapse.storage.events import EventsWorkerStore
+from synapse.storage.events_worker import EventsWorkerStore
 from synapse.storage.signatures import SignatureWorkerStore
 from synapse.util.caches.descriptors import cached
 
diff --git a/synapse/storage/events.py b/synapse/storage/events.py
index 2f482af3a1..c98e524ba1 100644
--- a/synapse/storage/events.py
+++ b/synapse/storage/events.py
@@ -34,6 +34,8 @@ from synapse.api.errors import SynapseError
 from synapse.events import EventBase  # noqa: F401
 from synapse.events.snapshot import EventContext  # noqa: F401
 from synapse.metrics.background_process_metrics import run_as_background_process
+from synapse.storage.background_updates import BackgroundUpdateStore
+from synapse.storage.event_federation import EventFederationStore
 from synapse.storage.events_worker import EventsWorkerStore
 from synapse.types import RoomStreamToken, get_domain_from_id
 from synapse.util.async import ObservableDeferred
@@ -193,7 +195,9 @@ def _retry_on_integrity_error(func):
     return f
 
 
-class EventsStore(EventsWorkerStore):
+# inherits from EventFederationStore so that we can call _update_backward_extremities
+# and _handle_mult_prev_events (though arguably those could both be moved in here)
+class EventsStore(EventFederationStore, EventsWorkerStore, BackgroundUpdateStore):
     EVENT_ORIGIN_SERVER_TS_NAME = "event_origin_server_ts"
     EVENT_FIELDS_SENDER_URL_UPDATE_NAME = "event_fields_sender_url"
 
diff --git a/synapse/storage/roommember.py b/synapse/storage/roommember.py
index 027bf8c85e..10dce21cea 100644
--- a/synapse/storage/roommember.py
+++ b/synapse/storage/roommember.py
@@ -24,7 +24,7 @@ from canonicaljson import json
 from twisted.internet import defer
 
 from synapse.api.constants import EventTypes, Membership
-from synapse.storage.events import EventsWorkerStore
+from synapse.storage.events_worker import EventsWorkerStore
 from synapse.types import get_domain_from_id
 from synapse.util.async import Linearizer
 from synapse.util.caches import intern_string
diff --git a/synapse/storage/stream.py b/synapse/storage/stream.py
index 25d0097b58..b9f2b74ac6 100644
--- a/synapse/storage/stream.py
+++ b/synapse/storage/stream.py
@@ -43,7 +43,7 @@ from twisted.internet import defer
 
 from synapse.storage._base import SQLBaseStore
 from synapse.storage.engines import PostgresEngine
-from synapse.storage.events import EventsWorkerStore
+from synapse.storage.events_worker import EventsWorkerStore
 from synapse.types import RoomStreamToken
 from synapse.util.caches.stream_change_cache import StreamChangeCache
 from synapse.util.logcontext import make_deferred_yieldable, run_in_background
diff --git a/synapse/util/caches/descriptors.py b/synapse/util/caches/descriptors.py
index f8a07df6b8..861c24809c 100644
--- a/synapse/util/caches/descriptors.py
+++ b/synapse/util/caches/descriptors.py
@@ -473,105 +473,101 @@ class CacheListDescriptor(_CacheDescriptorBase):
 
         @functools.wraps(self.orig)
         def wrapped(*args, **kwargs):
-            # If we're passed a cache_context then we'll want to call its invalidate()
-            # whenever we are invalidated
+            # If we're passed a cache_context then we'll want to call its
+            # invalidate() whenever we are invalidated
             invalidate_callback = kwargs.pop("on_invalidate", None)
 
             arg_dict = inspect.getcallargs(self.orig, obj, *args, **kwargs)
             keyargs = [arg_dict[arg_nm] for arg_nm in self.arg_names]
             list_args = arg_dict[self.list_name]
 
-            # cached is a dict arg -> deferred, where deferred results in a
-            # 2-tuple (`arg`, `result`)
             results = {}
-            cached_defers = {}
-            missing = []
+
+            def update_results_dict(res, arg):
+                results[arg] = res
+
+            # list of deferreds to wait for
+            cached_defers = []
+
+            missing = set()
 
             # If the cache takes a single arg then that is used as the key,
             # otherwise a tuple is used.
             if num_args == 1:
-                def cache_get(arg):
-                    return cache.get(arg, callback=invalidate_callback)
+                def arg_to_cache_key(arg):
+                    return arg
             else:
-                key = list(keyargs)
+                keylist = list(keyargs)
 
-                def cache_get(arg):
-                    key[self.list_pos] = arg
-                    return cache.get(tuple(key), callback=invalidate_callback)
+                def arg_to_cache_key(arg):
+                    keylist[self.list_pos] = arg
+                    return tuple(keylist)
 
             for arg in list_args:
                 try:
-                    res = cache_get(arg)
-
+                    res = cache.get(arg_to_cache_key(arg),
+                                    callback=invalidate_callback)
                     if not isinstance(res, ObservableDeferred):
                         results[arg] = res
                     elif not res.has_succeeded():
                         res = res.observe()
-                        res.addCallback(lambda r, arg: (arg, r), arg)
-                        cached_defers[arg] = res
+                        res.addCallback(update_results_dict, arg)
+                        cached_defers.append(res)
                     else:
                         results[arg] = res.get_result()
                 except KeyError:
-                    missing.append(arg)
+                    missing.add(arg)
 
             if missing:
+                # we need an observable deferred for each entry in the list,
+                # which we put in the cache. Each deferred resolves with the
+                # relevant result for that key.
+                deferreds_map = {}
+                for arg in missing:
+                    deferred = defer.Deferred()
+                    deferreds_map[arg] = deferred
+                    key = arg_to_cache_key(arg)
+                    observable = ObservableDeferred(deferred)
+                    cache.set(key, observable, callback=invalidate_callback)
+
+                def complete_all(res):
+                    # the wrapped function has completed. It returns a
+                    # a dict. We can now resolve the observable deferreds in
+                    # the cache and update our own result map.
+                    for e in missing:
+                        val = res.get(e, None)
+                        deferreds_map[e].callback(val)
+                        results[e] = val
+
+                def errback(f):
+                    # the wrapped function has failed. Invalidate any cache
+                    # entries we're supposed to be populating, and fail
+                    # their deferreds.
+                    for e in missing:
+                        key = arg_to_cache_key(e)
+                        cache.invalidate(key)
+                        deferreds_map[e].errback(f)
+
+                    # return the failure, to propagate to our caller.
+                    return f
+
                 args_to_call = dict(arg_dict)
-                args_to_call[self.list_name] = missing
+                args_to_call[self.list_name] = list(missing)
 
-                ret_d = defer.maybeDeferred(
+                cached_defers.append(defer.maybeDeferred(
                     logcontext.preserve_fn(self.function_to_call),
                     **args_to_call
-                )
-
-                ret_d = ObservableDeferred(ret_d)
-
-                # We need to create deferreds for each arg in the list so that
-                # we can insert the new deferred into the cache.
-                for arg in missing:
-                    observer = ret_d.observe()
-                    observer.addCallback(lambda r, arg: r.get(arg, None), arg)
-
-                    observer = ObservableDeferred(observer)
-
-                    if num_args == 1:
-                        cache.set(
-                            arg, observer,
-                            callback=invalidate_callback
-                        )
-
-                        def invalidate(f, key):
-                            cache.invalidate(key)
-                            return f
-                        observer.addErrback(invalidate, arg)
-                    else:
-                        key = list(keyargs)
-                        key[self.list_pos] = arg
-                        cache.set(
-                            tuple(key), observer,
-                            callback=invalidate_callback
-                        )
-
-                        def invalidate(f, key):
-                            cache.invalidate(key)
-                            return f
-                        observer.addErrback(invalidate, tuple(key))
-
-                    res = observer.observe()
-                    res.addCallback(lambda r, arg: (arg, r), arg)
-
-                    cached_defers[arg] = res
+                ).addCallbacks(complete_all, errback))
 
             if cached_defers:
-                def update_results_dict(res):
-                    results.update(res)
-                    return results
-
-                return logcontext.make_deferred_yieldable(defer.gatherResults(
-                    list(cached_defers.values()),
+                d = defer.gatherResults(
+                    cached_defers,
                     consumeErrors=True,
-                ).addCallback(update_results_dict).addErrback(
+                ).addCallbacks(
+                    lambda _: results,
                     unwrapFirstError
-                ))
+                )
+                return logcontext.make_deferred_yieldable(d)
             else:
                 return results
 
@@ -625,7 +621,8 @@ def cachedList(cached_method_name, list_name, num_args=None, inlineCallbacks=Fal
     cache.
 
     Args:
-        cache (Cache): The underlying cache to use.
+        cached_method_name (str): The name of the single-item lookup method.
+            This is only used to find the cache to use.
         list_name (str): The name of the argument that is the list to use to
             do batch lookups in the cache.
         num_args (int): Number of arguments to use as the key in the cache
diff --git a/tests/handlers/test_typing.py b/tests/handlers/test_typing.py
index b08856f763..2c263af1a3 100644
--- a/tests/handlers/test_typing.py
+++ b/tests/handlers/test_typing.py
@@ -44,7 +44,6 @@ def _expect_edu(destination, edu_type, content, origin="test"):
                 "content": content,
             }
         ],
-        "pdu_failures": [],
     }
 
 
diff --git a/tests/util/caches/test_descriptors.py b/tests/util/caches/test_descriptors.py
index 8176a7dabd..ca8a7c907f 100644
--- a/tests/util/caches/test_descriptors.py
+++ b/tests/util/caches/test_descriptors.py
@@ -273,3 +273,104 @@ class DescriptorTestCase(unittest.TestCase):
         r = yield obj.fn(2, 3)
         self.assertEqual(r, 'chips')
         obj.mock.assert_not_called()
+
+
+class CachedListDescriptorTestCase(unittest.TestCase):
+    @defer.inlineCallbacks
+    def test_cache(self):
+        class Cls(object):
+            def __init__(self):
+                self.mock = mock.Mock()
+
+            @descriptors.cached()
+            def fn(self, arg1, arg2):
+                pass
+
+            @descriptors.cachedList("fn", "args1", inlineCallbacks=True)
+            def list_fn(self, args1, arg2):
+                assert (
+                    logcontext.LoggingContext.current_context().request == "c1"
+                )
+                # we want this to behave like an asynchronous function
+                yield run_on_reactor()
+                assert (
+                    logcontext.LoggingContext.current_context().request == "c1"
+                )
+                defer.returnValue(self.mock(args1, arg2))
+
+        with logcontext.LoggingContext() as c1:
+            c1.request = "c1"
+            obj = Cls()
+            obj.mock.return_value = {10: 'fish', 20: 'chips'}
+            d1 = obj.list_fn([10, 20], 2)
+            self.assertEqual(
+                logcontext.LoggingContext.current_context(),
+                logcontext.LoggingContext.sentinel,
+            )
+            r = yield d1
+            self.assertEqual(
+                logcontext.LoggingContext.current_context(),
+                c1
+            )
+            obj.mock.assert_called_once_with([10, 20], 2)
+            self.assertEqual(r, {10: 'fish', 20: 'chips'})
+            obj.mock.reset_mock()
+
+            # a call with different params should call the mock again
+            obj.mock.return_value = {30: 'peas'}
+            r = yield obj.list_fn([20, 30], 2)
+            obj.mock.assert_called_once_with([30], 2)
+            self.assertEqual(r, {20: 'chips', 30: 'peas'})
+            obj.mock.reset_mock()
+
+            # all the values should now be cached
+            r = yield obj.fn(10, 2)
+            self.assertEqual(r, 'fish')
+            r = yield obj.fn(20, 2)
+            self.assertEqual(r, 'chips')
+            r = yield obj.fn(30, 2)
+            self.assertEqual(r, 'peas')
+            r = yield obj.list_fn([10, 20, 30], 2)
+            obj.mock.assert_not_called()
+            self.assertEqual(r, {10: 'fish', 20: 'chips', 30: 'peas'})
+
+    @defer.inlineCallbacks
+    def test_invalidate(self):
+        """Make sure that invalidation callbacks are called."""
+        class Cls(object):
+            def __init__(self):
+                self.mock = mock.Mock()
+
+            @descriptors.cached()
+            def fn(self, arg1, arg2):
+                pass
+
+            @descriptors.cachedList("fn", "args1", inlineCallbacks=True)
+            def list_fn(self, args1, arg2):
+                # we want this to behave like an asynchronous function
+                yield run_on_reactor()
+                defer.returnValue(self.mock(args1, arg2))
+
+        obj = Cls()
+        invalidate0 = mock.Mock()
+        invalidate1 = mock.Mock()
+
+        # cache miss
+        obj.mock.return_value = {10: 'fish', 20: 'chips'}
+        r1 = yield obj.list_fn([10, 20], 2, on_invalidate=invalidate0)
+        obj.mock.assert_called_once_with([10, 20], 2)
+        self.assertEqual(r1, {10: 'fish', 20: 'chips'})
+        obj.mock.reset_mock()
+
+        # cache hit
+        r2 = yield obj.list_fn([10, 20], 2, on_invalidate=invalidate1)
+        obj.mock.assert_not_called()
+        self.assertEqual(r2, {10: 'fish', 20: 'chips'})
+
+        invalidate0.assert_not_called()
+        invalidate1.assert_not_called()
+
+        # now if we invalidate the keys, both invalidations should get called
+        obj.fn.invalidate((10, 2))
+        invalidate0.assert_called_once()
+        invalidate1.assert_called_once()