[ https://issues.apache.org/jira/browse/SOLR-16577?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17842151#comment-17842151 ]
Houston Putman commented on SOLR-16577: --------------------------------------- [~dsmiley] I'm pretty sure we ran into issues upgrading to 9.5 that stem from this PR. The loading of cores asynchonously was simplified to remove the use of Futures to wait for the cores to be loaded. Instead, {{ExecutorUtil.shutdownAndAwaitTermination(coreLoadExecutor)}} was used to let the core loading start and complete. Unfortunately for us, {{ExecutorUtil.shutdownAndAwaitTermination(executor)}} defaults to a timeout of 60 seconds. This means that if your cores don't load in 60 seconds, then you will have problems. Notably, it will send interrupts to the coreLoading threads which will make it look like some of your configSet files cannot be found in ZK (which is quite an annoying red herring). We should either up this timeout value (should it really even timeout?) or return to using the methods that use "Future"s. > Core load issues are not always logged > -------------------------------------- > > Key: SOLR-16577 > URL: https://issues.apache.org/jira/browse/SOLR-16577 > Project: Solr > Issue Type: Improvement > Reporter: Haythem Khiri > Assignee: David Smiley > Priority: Minor > Fix For: 9.5 > > Time Spent: 2h 20m > Remaining Estimate: 0h > > It's possible for a core load failure to not have its cause logged. At least > the failure is tracked in a metric and one can do an admin request to fetch > the cause but really it ought to be logged so one can more easily see what > the problem is. -- This message was sent by Atlassian Jira (v8.20.10#820010) --------------------------------------------------------------------- To unsubscribe, e-mail: issues-unsubscr...@solr.apache.org For additional commands, e-mail: issues-h...@solr.apache.org