Guillaume Nodet created CAMEL-24020:
---------------------------------------
Summary: Fix flaky SjmsConnectionRecoveryTest
Key: CAMEL-24020
URL: https://issues.apache.org/jira/browse/CAMEL-24020
Project: Camel
Issue Type: Bug
Components: camel-sjms
Reporter: Guillaume Nodet
The SjmsConnectionRecoveryTest.testRecoveryStopsAfterSuccessfulReconnection
test is flaky on CI with a high failure rate (23 failures and 8 flaky runs out
of 271 total runs in the last 12 days).
The test uses asyncStartListener=true, which means the SJMS consumer starts
asynchronously in the background. The test sends messages immediately without
waiting for the consumer to be fully subscribed.
The flakiness was introduced by commit c5d06bf2e25 which removed Awaitility
wrapping from MockEndpoint assertions. While that removal was correct in
general (MockEndpoint already waits internally), in this specific case the
Awaitility wrapping served a different purpose: it retried the entire
send+assert cycle, which was necessary because:
1. Phase 1: the async consumer may not be subscribed when the first message is
sent, causing either a JMSException (Failed to create session factory) or the
message being lost (no consumer listening)
2. Phase 5: after recovery, the consumer needs time to fully re-establish
before it can process messages
The fix wraps the send+assert cycles with Awaitility.untilAsserted() to retry
until the consumer is ready, while keeping the direct
MockEndpoint.assertIsSatisfied() call inside the lambda for the actual
assertion.
--
This message was sent by Atlassian Jira
(v8.20.10#820010)
