wirybeaver commented on issue #10181: URL: https://github.com/apache/pinot/issues/10181#issuecomment-1405792829
Got that. Will use a new api for broker. "Force" is sort of misunderstanding. I haven't came up with an appropriate name yet. I actually wanted to do it in a safer way. Here is my implementation idea: (1) untag the broker first; (2) Then parallel call the rebuildBrokerResourceFromHelixTags api per table which is assigned to that node in the ideal state. (3) Delete the instance id. As for server, I suppose the same idea can be adopted. The step 2 can be replaced the existing table rebalance api. One concern the rebalancing is time consuming, shall we do it async? For example, we can set the state to "delete_pre" before step 1. And directly return a response (maybe a task id?) to user. Once the step 3 is done, set the state to "delete_complete". User can use this state (or the task id) to know whether the deletion is complete. If the long time workflow fail in the middle, maybe we can use minion (I suppose minion is good for task scheduling) to find out nodes with "delete_pre" state existing for >1h and retry. -- 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]
