C. Scott Andreas created CASSANDRA-21499:
--------------------------------------------
Summary: Flaky test: CatchupViaSnapshotTest#catchupViaSnapshotTest
Key: CASSANDRA-21499
URL: https://issues.apache.org/jira/browse/CASSANDRA-21499
Project: Apache Cassandra
Issue Type: Bug
Components: Transactional Cluster Metadata
Reporter: C. Scott Andreas
{{CatchupViaSnapshotTest.catchupViaSnapshotTest}} is intermittently flaky.
This test isolates node2/node3, drives 30 schema alters on node1 plus a {{{}cms
snapshot{}}}, then lets node2 catch up from node1 and node3 catch up from node2
via {{{}fetchLogFromPeerAsync{}}}. It finally asserted that every node's
{{ClusterMetadata}} epoch equalled node1's immediately after the async fetch
returned.
Because this convergence isn't instantaneous and there is no delay, a peer can
still be one epoch behind at the instant of the assertion. That produces a rare
{{junit.framework.AssertionFailedError: expected:<44> but was:<43>}} at line 70.
Polling for convergence with {{ClusterUtils.waitForCMSToQuiesce(cluster,
expectedEpoch)}} might be an approach to address this.
LLM-assisted details available at: [https://github.com/cscotta/cassandra/pull/6]
--
This message was sent by Atlassian Jira
(v8.20.10#820010)
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
