MultithreadedNativeStoreTest
Yes, this is a SHACL test that attempts to discover any issues with the parallel validation.
We’ve had problems with this test before. The last time it discovered that the use of java 8 parallel streams during validation was an issue. The test itself still uses parallel streams, so I might just rewrite it to use a pool instead to see if that helps.
Otherwise I know that there is a general problem with sharing the same connection between multiple threads (which is how parallel validation happens). Especially in the native store, since it uses a TreeModel for transaction isolation. The TreeModel does not support multithreaded reads (since it builds indexes in a lazy fashion).
Håvard
Hmz, seems that one of the (SHACL ?) tests in JDK11 took way too long, resulting in a build failure 😕
"
11:24:16.434 [main] DEBUG Nothing has changed, nothing to validate.
11:24:16.435 [main] DEBUG Cleanup
11:24:16.503 [ForkJoinPool.commonPool-worker-3] DEBUG Cleanup
11:24:16.585 [main] DEBUG Nothing has changed, nothing to validate.
11:24:16.586 [main] DEBUG Cleanup
11:24:16.610 [ForkJoinPool.commonPool-worker-3] DEBUG Cleanup
11:24:16.723 [ForkJoinPool.commonPool-worker-3] DEBUG Cleanup
11:24:16.988 [main] DEBUG Cleanup
"
Best regards,
Bart
Run failed for issues/#1580-vocab-dcat2 (2758f30)
Repository: eclipse/rdf4j
Workflow: PR unit test
Duration: 6 hours and 26.0 seconds
Finished: 2019-11-28 17:06:17 UTC
—
You are receiving this because this workflow ran on your branch.
Manage your GitHub Actions notifications
here.
![]()
