zhztheplayer commented on issue #12911: URL: https://github.com/apache/gluten/issues/12911#issuecomment-5790437227
> HookedExecutor::join() is not a thread join despite the name, it's a timeout, and expiry is only a log line. @boneanxs Do you have any clue which task is stocking on the IO thread pool? Based on the recent information including https://github.com/apache/gluten/pull/13038, I assume the default async-waiting timeout `spark.gluten.sql.columnar.backend.velox.asyncTimeoutOnTaskStopping=30s` is already reached? Which itself can cause significant performance issue already. You can set `spark.gluten.sql.debug=true`, then HookedExecutor will [print the timeout tasks](https://github.com/apache/gluten/blob/7e309eec6a5edfa2d3a4a37c80e238e4ee0b84c1/cpp/velox/compute/VeloxRuntime.cc#L108-L126) after the default 30s is reached. -- 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] --------------------------------------------------------------------- To unsubscribe, e-mail: [email protected] For additional commands, e-mail: [email protected]
