lizhimins opened a new issue, #11007:
URL: https://github.com/apache/rocketmq/issues/11007
### Before Creating the Bug Report
- [X] I found a bug, not just asking a question.
- [X] I have searched the GitHub Issues and Discussions of this repository
and believe that this is not a duplicate.
- [X] I have confirmed that this bug belongs to the current repository.
### Runtime platform environment
OS: Ubuntu 22.04 (GitHub Actions runner `ubuntu-latest`)
### RocketMQ version
branch: develop
version: 5.x
Git commit id: latest develop
### JDK Version
Compiler: JDK 8 (maven-compile job, ubuntu-latest)
### Describe the Bug
`CreateAndUpdateTopicIT` intermittently fails in the CI integration test
pipeline. The assertions on topic route size run immediately after topic
creation / broker registration, but route propagation from broker to NameServer
is asynchronous. Under CI load, the NameServer route table may not have been
fully updated when the assertion executes, causing a race condition.
Observed failure on an unrelated PR (apache/rocketmq#11001), CI run
33726673235:
```
CreateAndUpdateTopicIT.java#L142
Expected size: 3 but was: 2 in:
[BrokerData [brokerName=TestBrokerName_2, brokerAddrs={0=127.0.0.1:35821},
enableActingMaster=false],
BrokerData [brokerName=TestBrokerName_1, brokerAddrs={0=127.0.0.1:33565},
enableActingMaster=false]]
```
The failing test `testCreateOrUpdateTopic_EnableSplitRegistration` calls
`registerBrokerAll(...)` on all 3 brokers and then immediately asserts
`route.getBrokerDatas()).hasSize(3)`. One broker's registration had not yet
propagated to the NameServer when the assertion ran.
### Steps to Reproduce
1. Run the integration test `CreateAndUpdateTopicIT` in the CI pipeline (or
repeatedly under load).
2. The route-size assertions (`hasSize(3)`) in
`testCreateOrUpdateTopic_EnableSplitRegistration`,
`testCreateOrUpdateTopic_EnableSingleTopicRegistration`, and
`testStaticTopicNotAffected` execute without waiting for route propagation.
3. Occasionally the assertion observes only 2 BrokerData entries instead of
3.
Note: `testDeleteTopicFromNameSrvWithBrokerRegistration` in the same class
is already `@Ignore`d with the comment "Temporarily ignore the fact that this
test cannot pass in the integration test pipeline due to unknown reasons",
indicating this class has known flakiness.
### What Did You Expect to See?
The tests should wait for the route to propagate before asserting, and pass
deterministically.
### What Did You See Instead?
Intermittent failure: `Expected size: 3 but was: 2`.
### Additional Context
The fix is to wrap the route-size assertions in `awaitility` polling
(`await().atMost(30, TimeUnit.SECONDS).untilAsserted(...)`), consistent with
the existing pattern already used in
`testDeleteTopicFromNameSrvWithBrokerRegistration`.
--
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.
To unsubscribe, e-mail: [email protected]
For queries about this service, please contact Infrastructure at:
[email protected]
