[ https://issues.apache.org/jira/browse/CASSANDRA-19013?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel ]
Brandon Williams updated CASSANDRA-19013: ----------------------------------------- Resolution: Invalid Status: Resolved (was: Triage Needed) This jira is for the development of Apache Cassandra, and as such makes for a poor vehicle for support. We recommend contacting the community for assistance on the ML or slack instead: https://cassandra.apache.org/_/community.html If you are able to identify a specific defect to solve, then return to this jira with that information and we can re-open to take a look. > Performance issues after Cassandra upgrade > ------------------------------------------ > > Key: CASSANDRA-19013 > URL: https://issues.apache.org/jira/browse/CASSANDRA-19013 > Project: Cassandra > Issue Type: Bug > Components: Local/Config > Reporter: Márk Dezsőfi > Priority: Normal > > I would like to ask for your help in the following issue. > > We have upgraded our Cassandra 3.11.4 to 4.1.2 in production environment and > since that we're facing with many performance issues: > * The cluster operations lowered to 17-20k/s which was between 21-25k/s with > 3.11.4 > * CPU usage significantly increased -> previously it was nearly 50% at peak > times, after the update it's between 65-70% > * Pending tasks increased -> from nearly 0 to 30 at peak times > * Native Transport Requests significantly increased -> from 100-200 to 2K-3K > at peak times > * Chunck cache hit rate metric stopped working > * Read latency increased from microsecs to milisecs > * sstables per read histogram also shows increased values > We have already done with: > * file_cache_enabled has been enabled > * sstableupgrade > * Java has been updated from 8 to 11 > * Made a truncate of affected keyspace > * Affected keyspace's indexes has been rebuilt > Our infrastructure contains 2 DCs with 24 nodes/dc. Every node has 8 CPUs and > 64GBs of RAM and running on RHEL 8.6. > We have already compared the two versions but didn't notice any new feature > or option which could cause this performance degradation. If do you have any > idea or do you need any kind of logs don't hesitate to contact me. -- This message was sent by Atlassian Jira (v8.20.10#820010) --------------------------------------------------------------------- To unsubscribe, e-mail: commits-unsubscr...@cassandra.apache.org For additional commands, e-mail: commits-h...@cassandra.apache.org