keith-turner edited a comment on issue #2199: URL: https://github.com/apache/accumulo/issues/2199#issuecomment-879441650
The problem described in this issue can be [seen in a test](https://gist.github.com/keith-turner/29b72dfa97acad53c395dd0724b65f1c#results-for-1605-cluster) I ran a while ago. In this test user compactions were periodically scheduled while ingest was running. When the user compactions was initiated it immediately reserved all files on all tablet. However it would take a few hours for the user compaction to complete and a lot of tablets were queued for user compaction during this time with their reserved files unavailable to system compactions. This caused the average files per tablets to spike upward while the user compactions ran as seen in the metrics plots from the test. System compactions could still occur on non-reserved files that arrived after the user compaction was initiated for the table. I suspect if the change described in this issue were made that the files per tablet would not spike in the test (assuming user compactions and system compactions have different thread pools as they did in the test). -- 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]
