JVM freeze# Java - 爪哇娇娃
c*n
1 楼
the first impression I got is GC, but after looking at the print out from
-XX:+PrintGCDetails -XX:+PrintGCTimeStamps
I don't see a GC happening around the time of freeze.
basically it freezes for about 20 seconds, around this time, yourkit does
not take any samples, so it's a straight empty period of 20 seconds for all
threads on the yourkit threads display.
in a few cases it's indeed due to GC. but most other cases not: because I
opened sar , iostat and recorded system cpu, during most of the freezes, CPU
was actually low.
I could do a loop and catch thread dump whenever I see the CPU utilization
drop again, but I'm afraid it's not going to help me much. ----- what is
more, I did tcpdump, and clearly during the freeze window, I can see packets
going into and out of the box, so it's not a jvm freeze as in the sense of
stop-the-world GC.
any ideas of what this could be? and how to debug it?
Thanks
-XX:+PrintGCDetails -XX:+PrintGCTimeStamps
I don't see a GC happening around the time of freeze.
basically it freezes for about 20 seconds, around this time, yourkit does
not take any samples, so it's a straight empty period of 20 seconds for all
threads on the yourkit threads display.
in a few cases it's indeed due to GC. but most other cases not: because I
opened sar , iostat and recorded system cpu, during most of the freezes, CPU
was actually low.
I could do a loop and catch thread dump whenever I see the CPU utilization
drop again, but I'm afraid it's not going to help me much. ----- what is
more, I did tcpdump, and clearly during the freeze window, I can see packets
going into and out of the box, so it's not a jvm freeze as in the sense of
stop-the-world GC.
any ideas of what this could be? and how to debug it?
Thanks