[
https://issues.apache.org/jira/browse/FLINK-13531?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]
Chesnay Schepler closed FLINK-13531.
------------------------------------
Fix Version/s: 1.10.0
Resolution: Fixed
master: f9b7467c12d6f7f17d198fc4c9ee454c29739216
> Do not print log and call 'release' if no requests should be evicted from the
> shared slot
> -----------------------------------------------------------------------------------------
>
> Key: FLINK-13531
> URL: https://issues.apache.org/jira/browse/FLINK-13531
> Project: Flink
> Issue Type: Bug
> Components: Runtime / Coordination
> Affects Versions: 1.9.0
> Reporter: Yun Gao
> Assignee: Yun Gao
> Priority: Minor
> Labels: pull-request-available
> Fix For: 1.10.0
>
> Time Spent: 20m
> Remaining Estimate: 0h
>
> After adding the logic to bookkeeping the resource used in the shared slots,
> the resource requests will be recorded inside the MultiTaskSlot and when the
> underlying slot is allocated, all the resource requests will be checked if
> there is over-subscription, if so, some requests will be failed.
> In the current implementation, the code does not check the amount to fail
> before printing the over-allocated debug log and tries to fail them. This
> should not cause actual errors, but it will
> # Print a debug log saying some requests will be failed even if no one to
> fail.
> # If the total number of requests is 0 (This is possible if there already
> AllocatedSlot before the first request), the _release_ method will be called.
> Although it will do nothing with the current implementation (the slot is
> still being created and not added to any other data structure), it may cause
> error if the release logic changes in the future.
> To fix this issue, we should add a explicit check on the number of requests
> to fail.
>
--
This message was sent by Atlassian Jira
(v8.3.2#803003)
