[
https://issues.apache.org/jira/browse/NIFI-16562?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]
David Handermann updated NIFI-16562:
------------------------------------
Fix Version/s: 2.13.0
Resolution: Fixed
Status: Resolved (was: Patch Available)
> Fix flaky ClusteredStopSourcesIT and ClusteredConnectorAutoResumeIT
> -------------------------------------------------------------------
>
> Key: NIFI-16562
> URL: https://issues.apache.org/jira/browse/NIFI-16562
> Project: Apache NiFi
> Issue Type: Test
> Reporter: Pierre Villard
> Assignee: Pierre Villard
> Priority: Major
> Fix For: 2.13.0
>
> Time Spent: 1h 10m
> Remaining Estimate: 0h
>
> Several clustered system tests fail intermittently in CI for reasons
> unrelated to the changes under test.
> *
> {*}ClusteredStopSourcesIT{*}: Processors in a Stateless Process Group are
> started asynchronously on each node. The test waits on the cluster-replicated
> processor state, which does not merge the component state across nodes, so
> the wait can complete while one node still reports STOPPED. The test now
> waits on each node directly. A reusable {{waitForOnAllNodes}} helper is added
> to {{{}NiFiSystemIT{}}}.
> *
> {*}ClusteredConnectorAutoResumeIT{*}: A node restart failed with
> {{BindException}} on the cluster protocol port (48102). Ports 48101/48102
> fall within the default Linux ephemeral port range, so an outbound connection
> can take the port. The test cluster's protocol ports move to 6351/6352.
> Separately, the test framework kept waiting on a node that had already exited
> until the 5-minute test timeout. Startup now fails as soon as the node
> process exits, and {{AggregateNiFiInstance}} rethrows startup failures from
> its startup threads instead of dropping them.
--
This message was sent by Atlassian Jira
(v10.3.26#10030026)
