In the PR, there was a suggestion about running nightly tests on GitHub PRs (using Crave). To me it's kind of a smell that we have such poor feedback loops outside of PRs that we seek for PRs to be the ultimate in validation. If only we had better project notification mechanisms like a weekly email summarizing builds and recently failing tests. The data / reports are already in Develocity & Fucit.org.
On Tue, Jan 28, 2025 at 2:09 PM Houston Putman <hous...@apache.org> wrote: > After debugging locally, I think it might actually be from SOLR-17548. Will > try to come up with a fix. > > On Tue, Jan 28, 2025 at 9:52 AM Houston Putman <hous...@apache.org> wrote: > > > Hello everyone, > > > > Trying to get the tests working on a more consistent basis. An upgrade of > > mockito fixed the cross-dc failures that we were seeing on the Policeman > > boxes. Now it looks like "TestReplicationHandler" is our biggest failure. > > (100% on nightly main runs) > > > > It looks like this started out around December 20, 2024. (Or at least > > that's when the graph starts to spike, and when the last successful > nightly > > test run occurred) > > > > > http://fucit.org/solr-jenkins-reports/history-trend-of-recent-failures.html#series/org.apache.solr.handler.TestReplicationHandler.doTestRepeater > > > > SOLR-17306 <https://issues.apache.org/jira/browse/SOLR-17306> seems to > > have been committed on December 19th, so that looks like a good > candidate, > > but there was also a change on December 14th that touched the test and > the > > ReplicationHandler: SOLR-17579 > > <https://issues.apache.org/jira/browse/SOLR-17579>. Maybe it's one or a > > combination of those 2. > > > > - Houston > > >
