-
Type:
Task
-
Resolution: Unresolved
-
Priority:
Major - P3
-
None
-
Affects Version/s: None
-
Component/s: None
-
None
-
Replication, Replication - Oplog and Recovery
-
Repl 2026-11-23
-
0
-
None
-
None
-
None
-
None
-
None
-
None
-
None
In BF-47145, stepdown_fassert_on_explicit_overridden_rstl_timeout.js fails on 'windows-debug-suggested (replica_sets_auth)' with "assert.soon failed (timeout 600000ms), msg: waiting for state indicator" at the 'waitForState(primary, ReplSetTest.State.DOWN)' call.
The root cause is a race between the primary node asserting after stepdown and the following waitForState(primary, DOWN) call:
SERVER-137256wrapped asCluster()'s (jstests/libs/replsettest.js) connectionStatus call in retryOnNetworkError with 240 retries, 1s apart. The retries reuse the same connection and never go back to _callHello to pick another node.- This test calls waitForState(primary, DOWN) while the primary is still alive and about to crash. On a Windows variant, the primary node process keeps accepting connections for about 1s while it writes the crash dump, so asCluster()'s _callHello can still select the dying node as the connection to query.
- connectionStatus then fails once the primary fully shuts down, and all 240 retries run (takes about 13 minutes, which exceeds waitForState's 600s timeout).
Fix: Add a check after the stepDownCmd to wait until the primary no longer accepts connections, which proves the process has actually exited. Only then proceed to waitForState.
- is related to
-
SERVER-137256 retry connections when verifying repl dbhash
-
- Closed
-