[ https://issues.apache.org/jira/browse/SOLR-5397?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13824991#comment-13824991 ]
Mark Miller commented on SOLR-5397: ----------------------------------- I think I've got it - just have to test it out. We actually roll back the writer on the main index before we copy the files from the tmp index into - but that rollback can actually end up removing files - so we might have skipped some files thinking they were already in the std index (because they were), but then the rollback can remove them. This causes a corrupt index and opening a reader on it rolls back to the first empty commit. This has been around for some time, but I think recent changes allowed it to be exposed in tests. > Test fails for ChaosMonkeySafeLeaderTest on Apache Jenkins > ----------------------------------------------------------- > > Key: SOLR-5397 > URL: https://issues.apache.org/jira/browse/SOLR-5397 > Project: Solr > Issue Type: Bug > Components: SolrCloud > Reporter: Mark Miller > Assignee: Mark Miller > Fix For: 4.6, 5.0 > > Attachments: monkeyrun.log > > > Error Message: > shard3 is not consistent. Got 137 from > http://127.0.0.1:49317/collection1lastClient and got 125 from > http://127.0.0.1:39794/collection1 -- This message was sent by Atlassian JIRA (v6.1#6144) --------------------------------------------------------------------- To unsubscribe, e-mail: dev-unsubscr...@lucene.apache.org For additional commands, e-mail: dev-h...@lucene.apache.org