Summary

Java rarely crash (SIGSEGV) in HotSpot with error 'jvm.dll ciObjectFactory::create_new_metadata'

Environment

• Java 8u40+

Steps to Reproduce

Unknown

Expected Results

Works fine.

Actual Results

JVM crash, in hs_err_pidXX.log can be found:

# JRE version: Java(TM) SE Runtime Environment (8.0_45-b14) (build 1.8.0_45-b14)
# Java VM: Java HotSpot(TM) 64-Bit Server VM (25.45-b02 mixed mode linux-amd64 compressed oops)
# Problematic frame:
# V  [libjvm.so+0x432007]  ciObjectFactory::create_new_metadata(Metadata*)+0x327

---------------  T H R E A D  ---------------
Current thread (0x00007fcded5c2800):  JavaThread "C2 CompilerThread0" daemon [_thread_in_vm, id=27072, stack(0x00007fcd69e46000,0x00007fcd69f47000)]

siginfo: si_signo: 11 (SIGSEGV), si_code: 1 (SEGV_MAPERR), si_addr: 0x0000000000000013

Notes

See:

• https://bugs.openjdk.java.net/browse/JDK-8134389
• https://bugs.openjdk.java.net/browse/JDK-8078731

Outside of JIRA, we've observed Confluence running Java 8u40 and 8u66 being affected.

Workaround

For Jira 6.4 - downgrade to Java7

Given this is currently appearing to be a Java bug we're a bit limited in how we can help further.

Update 2016-02-09 from Oracle see [JDK-8134389

Status: IN PROGRESS
Understanding: Cause Known
Missing corner case handling of type 'instance_klass' in method LoadIndexed::declared_type. Similar corner case is taken care in another method “LoadIndexed::exact_type”.

Update 2016-02-12

• We see that in most cases there is -javaagent configured, but not sure if that 100% is related.
• Feedback from client:

  We upgraded to the New Relic 3.25.0 agent. Our Confluence is still on 5.8.4, and our Java is still 8u60. We got through the 24-hour danger zone without any nodes crashing with a SIGSEGV, so very promising!

Update 2017-05-29 from Oracle see [JDK-8134389

Status:RESOLVED
Fix Version/s: 9
Backport to Java 8u152, see https://bugs.openjdk.java.net/browse/JDK-8168123