LuciferYang commented on pull request #33748: URL: https://github.com/apache/spark/pull/33748#issuecomment-902378554
> Since the metadata is cached in the executor, does it mean the task reading the same ORC file has to be scheduled on the same executor? How can we guarantee this? At present, I think it is try best because there is no guarantee of scheduling, if the there has many `NODE_LOCAL` scheduling is probably the best scenario. We may need to collect the information about `fileMetaCache` distribution to the driver and consider this factor when task scheduling, but this may make the driver use more memory. -- 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]
