Not sure why this wasn't an issue before. Maybe Tomcat would kill these threads after so long, and since we are getting pummeled with queries, the threads didn't have a chance to die.
Either way, closing the solr client connections each time seems to fix it.
For the implementation, we could have opened it during the constructor, leave it open, and then have a closeConnection method but my concern there was we would have to add that close call to all the JSPs, and it is really hard to ensure those JSPs will complete execution to the bottom. Would rather have more open connections we know we close, versus leaving it to the client to close it.
Testing in production 😨
This thread count stayed stuck at 195 for 15 minutes while we had ~200 queries during that timeframe (196 when it was in the middle of a connection)
Fix memory leaks for solr connection left open.
Not sure why this wasn't an issue before. Maybe Tomcat would kill these threads after so long, and since we are getting pummeled with queries, the threads didn't have a chance to die.
Either way, closing the solr client connections each time seems to fix it.
For the implementation, we could have opened it during the constructor, leave it open, and then have a
closeConnectionmethod but my concern there was we would have to add that close call to all the JSPs, and it is really hard to ensure those JSPs will complete execution to the bottom. Would rather have more open connections we know we close, versus leaving it to the client to close it.Testing in production 😨
This thread count stayed stuck at 195 for 15 minutes while we had ~200 queries during that timeframe (196 when it was in the middle of a connection)
Fixes #14