Error: Shenandoah assert_correct failed; Object klass pointer must go to metaspace
Tom van der Woerdt
info at tvdw.eu
Fri Dec 4 18:34:20 UTC 2020
Heya,
I’m facing a crash when running Apache Cassandra under load, with Shenandoah on the Java 8 backport from RedHat. Also reproducible with the latest builds at https://builds.shipilev.net/openjdk-shenandoah-jdk8/
crash.txt, attached, shows the short version of the error. hs_err_pid42292.log has the longer version (from the fastdebug build).
I am able to consistently reproduce it in my environment, but it’s not exactly trivial. Steps @ https://gist.github.com/TvdW/0b8b6909c22a537363bb27217a706b1a
Some notes:
* -Xint stops the issue from happening
* -XX:TieredStopAtLevel=1 seems to make it more likely
* -XX:+ShenandoahVerify crashes by itself because of the same klass pointer issue
Since I’m quite certain this is a Sh bug, I’m sending it directly here; please let me know if there’s anything I can add to help figure this out. This is my first JVM bug so I’m not sure what you need from me ;-)
Tom
-------------- next part --------------
An embedded and charset-unspecified text was scrubbed...
Name: crash.txt
URL: <https://mail.openjdk.java.net/pipermail/shenandoah-dev/attachments/20201204/ea3036f3/crash.txt>
More information about the shenandoah-dev
mailing list