You are viewing a plain text version of this content. The canonical link for it is here.
Posted to issues@flink.apache.org by "Stephan Ewen (JIRA)" <ji...@apache.org> on 2019/07/26 07:42:00 UTC

[jira] [Resolved] (FLINK-12765) Bookkeeping of available resources of allocated slots in SlotPool.

     [ https://issues.apache.org/jira/browse/FLINK-12765?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel ]

Stephan Ewen resolved FLINK-12765.
----------------------------------
    Resolution: Fixed

Fixed in
  - 1.9.0 via 0b1dafd52d380a8c1b312fa5cf51185367cfd83d
  - master via dac02696a9bdb3ecdfd084b784a43e74ca3d9098

> Bookkeeping of available resources of allocated slots in SlotPool.
> ------------------------------------------------------------------
>
>                 Key: FLINK-12765
>                 URL: https://issues.apache.org/jira/browse/FLINK-12765
>             Project: Flink
>          Issue Type: Sub-task
>          Components: Runtime / Coordination
>    Affects Versions: 1.8.0, 1.9.0
>            Reporter: Xintong Song
>            Assignee: Yun Gao
>            Priority: Major
>              Labels: pull-request-available
>             Fix For: 1.9.0
>
>          Time Spent: 20m
>  Remaining Estimate: 0h
>
> In this version, a task will always requests slot with its own resource need. If the resource need is less than the default slot resource, it will always be allocated to a default sized slot. 
>  
> The extra resources in the slot leaves chances for other tasks within the same slot sharing group to fit in. To take these chance, SlotPool will maintain available resources of each allocated slot. Available resource of an allocated slot should always be the total resource of the slot minus resources of tasks already assigned onto the slot. In this way, the SlotPool would be able to determine whether another task can fit into the slot. If a task cannot fit into the slot, for slot sharing group the SlotPool should request another slot from the ResourceManager, and for colocation group it should fail the job.



--
This message was sent by Atlassian JIRA
(v7.6.14#76016)