DanielLeens commented on issue #12456: URL: https://github.com/apache/seatunnel/issues/12456#issuecomment-5892213012
Classification: A / Zeta class-loader lifecycle and Metaspace. The new run materially narrows the fault boundary. Its worker logs show both connector loaders being released, and the inspected `classLoaderCache` returns to zero after completion. That rules out an entry merely left in `DefaultClassLoaderService` for this run. Current `dev` removes a non-cache loader at reference count zero and clears matching thread context class loaders, so the surviving four `SeaTunnelChildFirstClassLoader` instances require a retaining path outside that cache entry. Please do not choose among individual Hadoop-cache cleanup, parent-first loading, or broad classloader redesign yet. The screenshots are useful context but do not provide a searchable retaining-edge record. Please paste the textual MAT Path to GC Roots for one post-release loader, including every strong-reference field/object edge and the retaining thread or static owner if one exists. Pair it with the final reference count and the same observation after one more identical job. That will distinguish a surviving static cache, a live thread or ThreadLocal, and another root without guessing. Cache mode for equal JAR sets remains a mitigation, not a root-cause fix. No label or assignment change in this pass. -- This is an automated message from the Apache Git Service. To respond to the message, please log on to GitHub and use the URL above to go to the specific comment. To unsubscribe, e-mail: [email protected] For queries about this service, please contact Infrastructure at: [email protected]
