Hi KristoffSC, Did you increase the parallelism of the vertex that has the largest parallelism? Or did you explicitly set tasks to be in different slot sharing group? With the default slot sharing, the number of slots required/used equals to the max parallelism of a JobVertex, which is 6 in your case.
KristoffSC <krzysiek.chmielew...@gmail.com> 于2020年1月9日周四 下午9:26写道： > Thank you David and Zhu Zhu, > this helps a lot. > > I have follow up questions though. > > Having this > /"Instead the Job must be stopped via a savepoint and restarted with a new > parallelism"/ > > and slot sharing [1] feature, I got the impression that if I would start my > cluster with more than 6 task slots, Flink will try deploy tasks across all > resources, trying to use all available resources during job submission > > I did a two tests having my original task. > 1. I started a Job Cluster with 7 task slots (7 task manager since in this > case 1 task manager has one task slot). > 2. I started a Session cluster with 28 task slots in total. In this case I > had 7 task managers, 4 task slot each. > > For case 1, I use "FLINK_JOB" variable as stated in [2]. For case 2, I > submitted my job from UI after Flink started to be operative. > > > For both cases it used only 6 task slots, so it was still reusing task > slots. I got the impression that it will try to use as much available > resources as it can. > > What do you think about this? > > > [1] > > https://ci.apache.org/projects/flink/flink-docs-release-1.9/concepts/runtime.html#task-slots-and-resources > [2] > > https://github.com/apache/flink/blob/release-1.9/flink-container/docker/README.md > > > > > > > > > -- > Sent from: > http://apache-flink-user-mailing-list-archive.2336050.n4.nabble.com/ >
