-
Bug
-
Resolution: Done
-
Major
-
5.5.2
-
None
-
False
-
-
False
-
-
When BasicTCP.enableSuspectEvents(true) is configured, graceful node departures can cause remaining cluster members to form singleton clusters.
Reproducer demonstrates this as:
=== Iteration 419: Coordinator node-1 leaving ===
[node-2] View changed: 3 -> 2 (delta=1)
2026-01-15T14:00:15.742252Z Connection.Receiver node-0 DEBUG node-0: connection closed by peer node-1 (IP=127.0.0.1:7801), sending up a suspect event
[node-0] View changed: 3 -> 2 (delta=1)
[node-0] View changed: 2 -> 1 (delta=1)
2026-01-15T14:00:15.758345Z Connection.Receiver node-2 DEBUG node-2: connection closed by peer node-0 (IP=127.0.0.1:7800), sending up a suspect event
Checking remaining nodes:
node-0: [node-0|840] (1) [node-0]
*** BUG REPRODUCED! Node node-0 formed singleton cluster!
node-2: [node-2|839] (2) [node-2, node-0]
This is tracked in WF as https://issues.redhat.com/browse/WFLY-21236
- blocks
-
WFLY-21236 Inconsistent topology after coordinator leaving when using TCP transport-based failure detection mechanism
-
- Coding In Progress
-
- is Informed by
-
JGRP-2967 Add test coverage for BasicTCP.enableSuspectEvents with graceful departures
-
- Resolved
-