ningyougang commented on issue #3272: improve loadbalancer's schedule algorithm URL: https://github.com/apache/incubator-openwhisk/pull/3272#issuecomment-364841239 @markusthoemmes , it seems `random` is not a good idea, in long time, it is `fairly`, but in a short time, `random` may not `failly`, may lead to some invoker nodes's action is more, but some others is `less`. How do i found the problem? Becuase i am testing the `zerodowntime deployment`, during `disable one invoker` short time, above problem exist. So i suggest to improve like below ``` invokerProgression .find(_._3 < invokerBusyThreshold) .orElse(invokerProgression.find(_._3 < invokerBusyThreshold * 2)) .orElse(invokerProgression.find(_._3 < invokerBusyThreshold * 3)) .orElse( if (invokerProgression.isEmpty) None else Some(invokerProgression.minBy(_._3))) .map(_._1) ``` through ` Some(invokerProgression.minBy(_._3)))`, when all invokers are overloaded, the lb will chose the least number of invoker node to execute. How do you think?
---------------------------------------------------------------- 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
