drcrallen opened a new pull request #6829: Don't force rebalance of the whole cluster all the time URL: https://github.com/apache/incubator-druid/pull/6829 When a cluster gets to sufficient size, it can take a while for the coordinator to finish its segment balancing rules. This PR adds a coordinator dynamic config option to allow a limit to the number of nodes which are considered for targets. It takes two samples of the servers, orders them by available size, and starts the balancing work trying to move segments from the fullest of the first sample, to the most empty of the second. CPU flame graph of coordinator in a real cluster. <img width="1711" alt="screen shot 2019-01-09 at 3 15 12 pm" src="https://user-images.githubusercontent.com/8213081/50935615-73282d80-1421-11e9-8a67-77fd4a267b17.png";> This feature is opt in. Prior behavior is mostly retained. The only change is that the From and the To list for rebalancing traverse in opposing directions with relation to available room for segments.
---------------------------------------------------------------- This is an automated message from the Apache Git Service. To respond to the message, please log on GitHub and use the URL above to go to the specific comment. For queries about this service, please contact Infrastructure at: [email protected] With regards, Apache Git Services --------------------------------------------------------------------- To unsubscribe, e-mail: [email protected] For additional commands, e-mail: [email protected]
![https://user-images.githubusercontent.com/8213081/50935615-73282d80-1421-11e9-8a67-77fd4a267b17.png"](https://user-images.githubusercontent.com/8213081/50935615-73282d80-1421-11e9-8a67-77fd4a267b17.png"){kind=link}