gaojun2048 commented on pull request #5113: URL: https://github.com/apache/incubator-dolphinscheduler/pull/5113#issuecomment-803812824
> > Need to call a database query for every command scheduling? This will cause the database to be very stressful, and you need to think about other ways. > > As for the feature, it meets the management needs of some people for the worker group. it will be welcomed by people. > > As for the tech, I think that the performance of database is just enough. And this scheme takes into account the development cost and maintainability, which is conducive to the development and optimization of newcomers. If the database query causes stresses, it seems that reading the data of all worker nodes maybe cause more stresses on zk every time. Please check ` registryCenter.getRegisterOperator().get(workerGroupPath + "/" + node)`. Of course, I will also consider optimizing database queries in the future pr. > > From the perspective of community development, we use good features to attract more users and developers, and iteratively improve our community and technology. So this is just the beginning, rather than the perfection of technical details. Isn't it a more interesting thing? Rome was not built in a day. Let's think back that the 1.2.x version of ds also fetched tasks directly from workers. Because the pressure of the database caused the concurrency of DS scheduling to be questioned, we have saved a lot of time in the latest version to reduce the pressure on the database. I still recommend optimizing the call to the database before the merge. -- 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]
