[jira] [Created] (FLINK-8066) Changed configuration of taskmanagers should recreate them
Stephen Gran created FLINK-8066: --- Summary: Changed configuration of taskmanagers should recreate them Key: FLINK-8066 URL: https://issues.apache.org/jira/browse/FLINK-8066 Project: Flink Issue Type: New Feature Reporter: Stephen Gran Priority: Minor When we redeploy the jobmanager to our mesos cluster with changed parameters affecting the taskmanagers (eg, change from 1 CPU per TM to 2 CPUs per TM), the existing taskmanagers are reused rather than replaced with new taskmanagers with new parameters. It seems like `recoverWorkers` in `org.apache.flink.mesos.runtime.clusterframework.MesosFlinkResourceManager` has most of the information it would need to be able to perform this convergence, and it doesn't seem like a large amount of work to do the check. My concern with starting to work on the issue there is that there may be a higher level, perhaps in `FlinkResourceManager` that should perform this work on both mesos and yarn. The two implementations look quite different, however, so this may be an over eager optimisation best left for later. I'm happy to look at a patch for this, but I wanted some input before starting the work to see where you thought this should live. -- This message was sent by Atlassian JIRA (v6.4.14#64029)
[jira] [Created] (FLINK-6408) Repeated loading of configuration files in hadoop filesystem code paths
Stephen Gran created FLINK-6408: --- Summary: Repeated loading of configuration files in hadoop filesystem code paths Key: FLINK-6408 URL: https://issues.apache.org/jira/browse/FLINK-6408 Project: Flink Issue Type: Bug Affects Versions: 1.2.1 Reporter: Stephen Gran Priority: Minor We are running flink on mesos in AWS. Checkpointing is enabled with an s3 backend, configured via the hadoop s3a filesystem implementation and done every second. We are seeing roughly 3 million log events per hour from a relatively small job, and it appears that this is because every s3 copy event reloads the hadoop configuration, which in turn reloads the flink configuration. The flink configuration loader is outputting each key/value pair every time it is invoked, leading to this volume of logs. While the logging is relatively easy to deal with - just a log4j setting - the behaviour is probably suboptimal. It seems that the configuration loader could easily be changed over to a singleton pattern to prevent the constant rereading of files. If you're interested, we can probably knock up a patch for this in a relatively short time. Cheers, -- This message was sent by Atlassian JIRA (v6.3.15#6346)
[jira] [Created] (FLINK-6336) Placement Constraints for Mesos
Stephen Gran created FLINK-6336: --- Summary: Placement Constraints for Mesos Key: FLINK-6336 URL: https://issues.apache.org/jira/browse/FLINK-6336 Project: Flink Issue Type: New Feature Components: Mesos Affects Versions: 1.2.0 Reporter: Stephen Gran Priority: Minor Fenzo supports placement constraints for tasks, and operators expose agent attributes to frameworks in the form of attributes about the agent offer. It would be extremely helpful in our multi-tenant cluster to be able to make use of this facility. -- This message was sent by Atlassian JIRA (v6.3.15#6346)