Merge pull request #6098 from matrix-org/erikj/cleanup_user_ips_2

Prune old rows in user_ips tables.

1 files changed, 54 insertions, 8 deletions

diff --git a/synapse/storage/client_ips.py b/synapse/storage/client_ips.py
index 8996689744..539584288d 100644
--- a/synapse/storage/client_ips.py
+++ b/synapse/storage/client_ips.py
@@ -19,7 +19,7 @@ from six import iteritems
 
 from twisted.internet import defer
 
-from synapse.metrics.background_process_metrics import run_as_background_process
+from synapse.metrics.background_process_metrics import wrap_as_background_process
 from synapse.util.caches import CACHE_SIZE_FACTOR
 
 from . import background_updates
@@ -42,6 +42,8 @@ class ClientIpStore(background_updates.BackgroundUpdateStore):
 
         super(ClientIpStore, self).__init__(db_conn, hs)
 
+        self.user_ips_max_age = hs.config.user_ips_max_age
+
         self.register_background_index_update(
             "user_ips_device_index",
             index_name="user_ips_device_id",
@@ -100,6 +102,9 @@ class ClientIpStore(background_updates.BackgroundUpdateStore):
             "before", "shutdown", self._update_client_ips_batch
         )
 
+        if self.user_ips_max_age:
+            self._clock.looping_call(self._prune_old_user_ips, 5 * 1000)
+
     @defer.inlineCallbacks
     def _remove_user_ip_nonunique(self, progress, batch_size):
         def f(conn):
@@ -319,20 +324,19 @@ class ClientIpStore(background_updates.BackgroundUpdateStore):
 
         self._batch_row_update[key] = (user_agent, device_id, now)
 
+    @wrap_as_background_process("update_client_ips")
     def _update_client_ips_batch(self):
 
         # If the DB pool has already terminated, don't try updating
         if not self.hs.get_db_pool().running:
             return
 
-        def update():
-            to_update = self._batch_row_update
-            self._batch_row_update = {}
-            return self.runInteraction(
-                "_update_client_ips_batch", self._update_client_ips_batch_txn, to_update
-            )
+        to_update = self._batch_row_update
+        self._batch_row_update = {}
 
-        return run_as_background_process("update_client_ips", update)
+        return self.runInteraction(
+            "_update_client_ips_batch", self._update_client_ips_batch_txn, to_update
+        )
 
     def _update_client_ips_batch_txn(self, txn, to_update):
         if "user_ips" in self._unsafe_to_upsert_tables or (
@@ -496,3 +500,45 @@ class ClientIpStore(background_updates.BackgroundUpdateStore):
             yield self._end_background_update("devices_last_seen")
 
         return updated
+
+    @wrap_as_background_process("prune_old_user_ips")
+    async def _prune_old_user_ips(self):
+        """Removes entries in user IPs older than the configured period.
+        """
+
+        if self.user_ips_max_age is None:
+            # Nothing to do
+            return
+
+        if not await self.has_completed_background_update("devices_last_seen"):
+            # Only start pruning if we have finished populating the devices
+            # last seen info.
+            return
+
+        # We do a slightly funky SQL delete to ensure we don't try and delete
+        # too much at once (as the table may be very large from before we
+        # started pruning).
+        #
+        # This works by finding the max last_seen that is less than the given
+        # time, but has no more than N rows before it, deleting all rows with
+        # a lesser last_seen time. (We COALESCE so that the sub-SELECT always
+        # returns exactly one row).
+        sql = """
+            DELETE FROM user_ips
+            WHERE last_seen <= (
+                SELECT COALESCE(MAX(last_seen), -1)
+                FROM (
+                    SELECT last_seen FROM user_ips
+                    WHERE last_seen <= ?
+                    ORDER BY last_seen ASC
+                    LIMIT 5000
+                ) AS u
+            )
+        """
+
+        timestamp = self.clock.time_msec() - self.user_ips_max_age
+
+        def _prune_old_user_ips_txn(txn):
+            txn.execute(sql, (timestamp,))
+
+        await self.runInteraction("_prune_old_user_ips", _prune_old_user_ips_txn)