2

JVM gurus, Need your help. I have a JVM/app that is getting into a "suspended" state. It sounds similar to this older thread: How does jstack -F affect a running Java process?

...but I'm looking for any further suggestions/guidance.

  • its a commercial java app that (when seemingly idle - middle of the night, etc) becomes unresponsive.
  • "kill -3 " hangs, no response in console.
  • verbose gc log won't be written to at all.
  • 'top' shows pid having low (almost no) CPU or memory utilization, low file descriptors, etc.
  • jvm remains in this state indefinitely (we left it for 10 hrs once), until...
  • "jstack -F " is issued, which kickstarts the JVM/app. GC logging resumes, application starts responding again, etc.
  • issue has been occurring for a few weeks, near daily, to 30 independent JVMs (spread across 10 RHEL 6.6 VMs).
  • sometimes it happens in a matter of hours after startup. App can be completely unused during that time.
  • we've have been using Oracle JVM 1.8.0_31 for Linux x64 to date. Have just upgraded to latest public (1.8.0_72) and will see if issue goes away.

Anyone pointers? Thanks.

Community
  • 1
  • 1
MikeO
  • 23
  • 5

1 Answers1

3

You're likely hit by the futex_wait bug.

Update to a kernel that contains the fix, rev 76835b0e

the8472
  • 40,999
  • 5
  • 70
  • 122
  • Thanks the8472! We had not heard of this before, but it does look like our issue...we've upgraded some of the OSes to RHL 6.7 and are monitoring. – MikeO Feb 04 '16 at 02:59
  • FYI - after upgrading RHL to 6.7, issue was not seen for 10+ days. Thanks again the8472. – MikeO Feb 17 '16 at 21:49