DanielLeens commented on issue #12456: URL: https://github.com/apache/seatunnel/issues/12456#issuecomment-5794529186
Classification: A / Zeta runtime memory (Metaspace). Thanks for the report. The evidence currently establishes that the worker reaches the configured 2 GiB Metaspace ceiling while the Hive sink factory is being created; it does not yet identify which class loader, generated class, or repeatedly loaded artifact is retaining the metadata. The large Java heap settings do not diagnose a Metaspace exhaustion. Could you add a short English or bilingual summary (to keep the report searchable), then provide the following from the affected worker, with credentials and private host names redacted: 1. the full worker log from process start to the first OOM, plus the matching master log; 2. `java -version`, the effective JVM flags, and whether workers/jobs are restarted or re-submitted repeatedly before the failure; 3. `jcmd <pid> VM.classloader_stats` and `jcmd <pid> GC.class_histogram` shortly before/after the growth, or the generated heap dump if it is available; and 4. the installed connector/plugin JAR inventory and whether the failure happens only when a Hive sink is present. Those artifacts distinguish a class-loader retention issue from a Hive/Hadoop dependency-loading problem or a deployment-level repeated-JAR load. Please do not post secrets, keytabs, or full production configuration. Once we have that evidence, we can reproduce the narrow path and decide whether a code fix is needed. -- 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]
