ivakegg opened a new issue #1621: URL: https://github.com/apache/accumulo/issues/1621
The configuration parameter is tserver.server.thread.minimum. However the tserver (using the SimpleThreadPool) uses that as the core and max threads AND allows core threads to time out effectively allowing the number of waiting threads to fall to 0. A timer is created in the createSelfResizingThreadPool method of TServerUtils which will move the max and core size up and down (but not below tserver.server.thread.minimum) depending on the number of active threads relative to the core threads. This normally works just fine and appears to adhere to the original design (see [this ACCUMULO-669 commit](https://github.com/apache/accumulo/commit/2b3bfde11cd861c1d39504684ae516070069b0f9)). However when systems get large and have bursty activities (e.g. bulk loads, major compaction storms, etc.), then we have seem tservers that end up spending all of their time starting and stopping threads when handling metadata tablets. It would be nice to be able to configure tservers to NOT allow core threads to time out or alternately allow configuration of the timeout value. ---------------------------------------------------------------- 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. For queries about this service, please contact Infrastructure at: [email protected]
