[ https://issues.apache.org/jira/browse/FLINK-1663?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=14353312#comment-14353312 ]
Robert Metzger commented on FLINK-1663: --------------------------------------- Yep, users often request this. +1 > Add a FIFO job scheduler to the JobManager > ------------------------------------------ > > Key: FLINK-1663 > URL: https://issues.apache.org/jira/browse/FLINK-1663 > Project: Flink > Issue Type: New Feature > Components: JobManager > Reporter: Fabian Hueske > Priority: Minor > > Right now, the JobManager rejects jobs that cannot be immediately executed > due to lack of available slots. > In case some slots are occupied by currently running jobs, it would be good > if the JM would put the job in a FIFO and execute it as soon as enough > resources became available. The scheduler should also monitor the total > amount of slots such that a job does not become a blocker if one or more TMs > go down. > A nice addition could be the option to define job interdependencies, i.e., a > job does only execute if some other job has been successfully executed. If > that other job failed, the job is dropped as well. -- This message was sent by Atlassian JIRA (v6.3.4#6332)