marek cervenka ?????: > hi, > > i'm testing asterisk 1.6/1.4/1.2 in the following scenario > centos5/cpu quad xeon E5335 2.0Ghz > - test clients behind nat > - 1500+ testing instances - reregister option from 1min to 1hour > - qualify set to 5000 > > top shows over 100% cpu. cpu cores sometimes go to 95% > with htop i see ~16threads but only one child have ~95% cpu > (how i can get info about that thread? what he is doing?) >
You need to use a software called "Profiler" there are at least 3 profilers for linux: gprof, oprofile and callgrind from valgrind package. I wasn't able to run gprof with asterisk, there is something with asterisk ending that don't fit gprof.. oprofile works nice though.. > what is major bottleneck? > qualify imho not. i'm tried set qualify=no, > this does not help > SIP REGISTER packets? > > this problem persist if no calls are active > after restart cpu usage slowly increase. after a ~hour is about 100% > > which optimalizations do you recommend for ~1500 peers scenario? (behind > nat, reregistrations) > how i can debug which thread/operations are most cpu intesive? strace? > gdb? can you show some examples? > > thanks > > --------------------------------------- > Marek Cervenka > ======================================= > > _______________________________________________ > --Bandwidth and Colocation Provided by http://www.api-digital.com-- > > asterisk-dev mailing list > To UNSUBSCRIBE or update options visit: > http://lists.digium.com/mailman/listinfo/asterisk-dev > _______________________________________________ --Bandwidth and Colocation Provided by http://www.api-digital.com-- asterisk-dev mailing list To UNSUBSCRIBE or update options visit: http://lists.digium.com/mailman/listinfo/asterisk-dev
