Weblogic 12 JDK 1.7 "jvmtiStringPrimitiveCallback should not be invoked when string value is null"
HI,
One of our clients is experiencing the following error "jvmtiStringPrimitiveCallback" before the JVM crashes with a Fatal Java Error.
This has been reported before on the openjdk forum (http://comments.gmane.org/gmane.comp.java.openjdk.hotspot.devel/12363) and Dynatrace has been identified as a possible cause for this.
JVM Information:
Weblogic 12 64 bit (Linux)
JDK 1.7
I will raise a support ticket for this but just curious if anyone else has experienced this problem?
Thanks
Richard
Best Answer
My apologies we did actually raise a support ticket for this and It appears to have been caused by the automatic memory dumps that include string capturing by default. We disabled automatic memory dumps on the agent group configuration which resolved the issue. Also in the longer term the issue was resolved by increasing the heap memory for the JVM's as they were having memory issues so maybe check if you aren't having memory problems on the jvm's?
Hi Richard,
We are also experiencing the same issue, Weblogic 12c and JDK 7. Our logs show:
# Problematic frame:
# V [libjvm.so+0x708d5b] invoke_string_value_callback(int (*)(long, long, long*, unsigned short const*, int, void*), CallbackWrapper*, oopDesc*, void*)+0x11b
How is dynaTrace progressing on finding a permanent fix? We have disabled automated memory dumps temporarily
Thanks,
Kevin
Thank you so much for your quick replies! I am fairly certain we are having Memory issues on this JVM, however, we have increased heap size previously- only to have the OutOfMemory condition return. So, it's vital that we get a heap dump to analyze. One thing I found while researching this is a "bug" caused by a method invocation common among profiling tools:
"# Problematic frame:
# V [libjvm.so+0x708d0b] invoke_string_value_callback(int (*)(long, long, long*, unsigned short const*, int, void*), CallbackWrapper*, oopDesc*, void*)+0x11b
https://bugs.openjdk.java.net/browse/JDK-6584008
One of the profiler vendors reports a crash when they enable the jvmtiStringPrimitiveCallback implementation. From the description it appears that the heap iteration (IterateThroughHeap) encounters a partially constructed java.lang.String. That is, the value field is null. The invoke_string_value_callback function doesn't check for this case and so will crash."
So, I want to see if this is something known in dynatrace- and if it's been addressed specifically.
We are opening a case today-
Thanks again!
Hi Philip, we didn't raise a support ticket for it the client rather removed dynatrace for a while to rule it out. However they did find the cause and fixed it I will find out what they did and get back to you.
Richard
Hi Richard, we are suspecting that this issue happened in our Jetty Environment- did you open a case and get any feedback?
