wilfred-s opened a new pull request #253: URL: https://github.com/apache/incubator-yunikorn-core/pull/253
When a node gets removed the node alocations get removed as part of the cleanup phase. The node removal acquires the partition lock and does not release it until all actions are done. This could cause a deadlock with scheduling if the node being removed contains allocations for the same application that is scheduled. The node removal should only lock while updating the partition and release before updating the apps and queues. Since node removal is called as part of the cleanup of a failed add of a node the add flow is changed inline with the removal to minimise lock time. The partition lock was also held while placing applications. This is not required as placing an application does not modify the partition until everything is done. The lock should only be acquired at that point. ---------------------------------------------------------------- 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]
