Allow config of the backoff algorithm for the federation client. (#15754)

Adds three new configuration variables:

* destination_min_retry_interval is identical to before (10mn).
* destination_retry_multiplier is now 2 instead of 5, the maximum value will
  be reached slower.
* destination_max_retry_interval is one day instead of (essentially) infinity.

Capping this will cause destinations to continue to be retried sometimes instead
of being lost forever. The previous value was 2 ^ 62 milliseconds.

2 files changed, 18 insertions, 13 deletions

diff --git a/tests/storage/test_transactions.py b/tests/storage/test_transactions.py
index 2fab84a529..ef06b50dbb 100644
--- a/tests/storage/test_transactions.py
+++ b/tests/storage/test_transactions.py
@@ -17,7 +17,6 @@ from twisted.test.proto_helpers import MemoryReactor
 from synapse.server import HomeServer
 from synapse.storage.databases.main.transactions import DestinationRetryTimings
 from synapse.util import Clock
-from synapse.util.retryutils import MAX_RETRY_INTERVAL
 
 from tests.unittest import HomeserverTestCase
 
@@ -57,8 +56,14 @@ class TransactionStoreTestCase(HomeserverTestCase):
         self.get_success(d)
 
     def test_large_destination_retry(self) -> None:
+        max_retry_interval_ms = (
+            self.hs.config.federation.destination_max_retry_interval_ms
+        )
         d = self.store.set_destination_retry_timings(
-            "example.com", MAX_RETRY_INTERVAL, MAX_RETRY_INTERVAL, MAX_RETRY_INTERVAL
+            "example.com",
+            max_retry_interval_ms,
+            max_retry_interval_ms,
+            max_retry_interval_ms,
         )
         self.get_success(d)
 
diff --git a/tests/util/test_retryutils.py b/tests/util/test_retryutils.py
index 5f8f4e76b5..1277e1a865 100644
--- a/tests/util/test_retryutils.py
+++ b/tests/util/test_retryutils.py
@@ -11,12 +11,7 @@
 # WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
 # See the License for the specific language governing permissions and
 # limitations under the License.
-from synapse.util.retryutils import (
-    MIN_RETRY_INTERVAL,
-    RETRY_MULTIPLIER,
-    NotRetryingDestination,
-    get_retry_limiter,
-)
+from synapse.util.retryutils import NotRetryingDestination, get_retry_limiter
 
 from tests.unittest import HomeserverTestCase
 
@@ -42,6 +37,11 @@ class RetryLimiterTestCase(HomeserverTestCase):
 
         limiter = self.get_success(get_retry_limiter("test_dest", self.clock, store))
 
+        min_retry_interval_ms = (
+            self.hs.config.federation.destination_min_retry_interval_ms
+        )
+        retry_multiplier = self.hs.config.federation.destination_retry_multiplier
+
         self.pump(1)
         try:
             with limiter:
@@ -57,7 +57,7 @@ class RetryLimiterTestCase(HomeserverTestCase):
         assert new_timings is not None
         self.assertEqual(new_timings.failure_ts, failure_ts)
         self.assertEqual(new_timings.retry_last_ts, failure_ts)
-        self.assertEqual(new_timings.retry_interval, MIN_RETRY_INTERVAL)
+        self.assertEqual(new_timings.retry_interval, min_retry_interval_ms)
 
         # now if we try again we should get a failure
         self.get_failure(
@@ -68,7 +68,7 @@ class RetryLimiterTestCase(HomeserverTestCase):
         # advance the clock and try again
         #
 
-        self.pump(MIN_RETRY_INTERVAL)
+        self.pump(min_retry_interval_ms)
         limiter = self.get_success(get_retry_limiter("test_dest", self.clock, store))
 
         self.pump(1)
@@ -87,16 +87,16 @@ class RetryLimiterTestCase(HomeserverTestCase):
         self.assertEqual(new_timings.failure_ts, failure_ts)
         self.assertEqual(new_timings.retry_last_ts, retry_ts)
         self.assertGreaterEqual(
-            new_timings.retry_interval, MIN_RETRY_INTERVAL * RETRY_MULTIPLIER * 0.5
+            new_timings.retry_interval, min_retry_interval_ms * retry_multiplier * 0.5
         )
         self.assertLessEqual(
-            new_timings.retry_interval, MIN_RETRY_INTERVAL * RETRY_MULTIPLIER * 2.0
+            new_timings.retry_interval, min_retry_interval_ms * retry_multiplier * 2.0
         )
 
         #
         # one more go, with success
         #
-        self.reactor.advance(MIN_RETRY_INTERVAL * RETRY_MULTIPLIER * 2.0)
+        self.reactor.advance(min_retry_interval_ms * retry_multiplier * 2.0)
         limiter = self.get_success(get_retry_limiter("test_dest", self.clock, store))
 
         self.pump(1)