-
Type:
Task
-
Resolution: Fixed
-
Priority:
Major - P3
-
Affects Version/s: None
-
Component/s: None
-
None
-
Catalog and Routing
-
Fully Compatible
-
CAR Team 2026-07-20
-
200
-
🟦 Shard Catalog
-
None
-
None
-
None
-
None
-
None
-
None
Problem
ContinuousAddRemoveShard and other resmoke hooks retry transient network errors using an allowlist in buildscripts/resmokelib/testing/retry.py. BF-44760 showed the hook aborting on ConnectionClosedByPeer instead of retrying during stepdown/kill-primary burn-in tests.
Root Cause
Commit 2ba57ec85427 (SERVER-123418 Limit pool-flush suppression in finishRefresh to confirmed rate-limiter signals (#52989)) introduced two new retryable network errors, ConnectionClosedByPeer (515) and ConnectionEstablishmentTimeout (516), for cases that previously surfaced as HostUnreachable (6). The resmoke allowlist was not updated.
Fix
Add 515 and 516 to retryable_codes in buildscripts/resmokelib/testing/retry.py so hooks retry these errors like the HostUnreachable cases they replaced.
Related
- BF-44760
- Commit 2ba57ec85427 (
SERVER-123418)
- is duplicated by
-
SERVER-131736 Overly strict retry predicate in ContinuousAddRemoveShard
-
- Closed
-
- is related to
-
SERVER-123418 Limit conditions in which connection pool avoids clearing in response to establishment failures
-
- Closed
-