It looks like a memory leak to me. All Cassandra processes are with the following JVM settings configured: -Xms10240M -Xmx10240M -Xmn2400M. The OldGen reaches nearly the maximum and it looks like the oldgen objects not being garbage collected.
Alexey On Fri, May 23, 2014 at 3:56 PM, Nate McCall <n...@thelastpickle.com> wrote: > Is compaction keeping up? Those replication settings are high - for every > write, 10 nodes are doing something). > > What other monitoring stats do you have - what is IO, CPU and network > traffic like? Is the JVM GC activity growing? > > Anything else stick out like growing number of network connections to 9160 > or 9042 on the cluster? > > The only leak-y things I've seen lately could either be: > > from a JVM bug leaking permgen from JMX invocation: > https://issues.apache.org/jira/browse/CASSANDRA-6541 > > or if you have anything using thrift (this includes opscenter) and you are > using HSHA as the rpc_server_type: > https://issues.apache.org/jira/browse/CASSANDRA-4265 > > Otherwise, I think this is config/setup tuning. > > > > > On Fri, May 23, 2014 at 7:00 AM, Alexey <alexe...@gmail.com> wrote: > >> Hi all, >> >> I've noticed increased latency on our tomcat REST-service (average 30ms, >> max > 2sec). We are using Cassandra 1.2.16 with official DataStax Java >> driver v1.0.3. >> >> Our setup: >> >> * 2 DCs >> * each DC: 7 nodes >> * RF=5 >> * Leveled compaction >> >> After cassandra restart on all nodes, the latencies are alright again >> (average < 5ms, max 50ms). >> >> Any thoughts are greatly appreciated. >> >> Thanks, >> Alexey >> > > > > -- > ----------------- > Nate McCall > Austin, TX > @zznate > > Co-Founder & Sr. Technical Consultant > Apache Cassandra Consulting > http://www.thelastpickle.com >