RFR: 8354894: java/lang/Thread/virtual/Starvation.java timeout on server with high CPUs

Alan Bateman alanb at openjdk.org
Tue Sep 30 07:30:17 UTC 2025


On Mon, 29 Sep 2025 08:52:04 GMT, SendaoYan <syan at openjdk.org> wrote:

> Hi all,
> 
> We observed java/lang/Thread/virtual/Starvation.java intermittent timed out on server with many CPU core number. The more CPU core number the 'runTest' will create more pinned virtual threads. This PR compute an iteration count which divide the CPU core number will make test run passed steady on server with high CPUs.
> 
> Before apply the proposed patch, test run finish 257 seconds, after apply the proposed patch test run finish 1.2 seconds on the 2 sockets system with Intel(R) Xeon(R) Platinum 8480+. So I remove the 'timeout=800', I think the default timeout 120 will be enough.

Thanks for taking this one. This looks like the patch I added to the JBS issue so I can't add myself as reviewer, hopefully someone else can jump in. There is another "Starvaration" test in the forkjoin directory that will also catch an under signal so I think it's okay to dial down the virtual/Starvation.java test.

-------------

PR Comment: https://git.openjdk.org/jdk/pull/27543#issuecomment-3350381723


More information about the core-libs-dev mailing list