Thanks for input Ivan, I'll keep that in mind in case we experience something similar again.
Cheers, Jens On Wed, Feb 24, 2016 at 3:56 PM, Ivan Kelly <[email protected]> wrote: > > Unfortunately, we are running Zookeeper with OpenJDK at the moment so I > was > > unable to execute `jstack` (it's not bundled). If this issues remains we > > might migrate to use Oracle's JDK instead. > If jstack is unavailable, you can get the stack printed to stdout with > kill -QUIT <pid>. The unavailability of jstack isn't a Oracle/OpenJDK > thing, but more likely that you only have the JRE installed, not the > JDK. > > Still, the stacks won't necessarily tell you what is eating CPU. > Do a "ps -eLo pid,ppid,lwp,nlwp,osz,rss,ruser,pcpu,stime,etime,args" > which will give you cpu usage per thread (lwp). The lwp will correlate > with the nxid in the stack output (though one is decimal, the other > hex). This should lead you to the thread that's eating your cpu (I'm > going to bet it's GC). > > -Ivan > -- Jens Rantil Backend engineer Tink AB Email: [email protected] Phone: +46 708 84 18 32 Web: www.tink.se Facebook <https://www.facebook.com/#!/tink.se> Linkedin <http://www.linkedin.com/company/2735919?trk=vsrp_companies_res_photo&trkInfo=VSRPsearchId%3A1057023381369207406670%2CVSRPtargetId%3A2735919%2CVSRPcmpt%3Aprimary> Twitter <https://twitter.com/tink>
