[
https://issues.apache.org/jira/browse/FLINK-32254?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17729597#comment-17729597
]
Weijie Guo edited comment on FLINK-32254 at 6/6/23 6:03 AM:
------------------------------------------------------------
[~huwh] In fact, the {{defaultWorkerResourceSpec.getTotalMemSize()}} is used to
calculate {{maxTotalMem}}, which is not exactly equal to
{{taskmanager.memory.process.size}}.
The actual ResourceProfile is as follows:
defaultSlotResourceProfile=ResourceProfile {CPUCores=1, taskHeapMemory=1.084gb
(1164338772 bytes), taskOffHeapMemory=0 bytes, managed Memory=908.800mb
(952945883 bytes), networkMemory=227.20mb (238236470 bytes).
was (Author: weijie guo):
[~huwh] In fact, the {{defaultWorkerResourceSpec.getTotalMemSize()}} is used to
calculate {{maxTotalMem}}, which is not exactly equal to
{{taskmanager.memory.process.size}}.
> FineGrainedSlotManager may not allocate enough taskmangers if maxSlotNum is
> configured
> --------------------------------------------------------------------------------------
>
> Key: FLINK-32254
> URL: https://issues.apache.org/jira/browse/FLINK-32254
> Project: Flink
> Issue Type: Bug
> Components: Runtime / Coordination
> Affects Versions: 1.18.0
> Reporter: Weijie Guo
> Assignee: Weijie Guo
> Priority: Major
> Labels: pull-request-available
>
> We ran a job with {{slotmanager.number-of-slots.max = 10}},
> {{taskmanager.numberOfTaskSlots = 10}} and {{taskmanager.memory.process.size:
> 24000m}}. The resources of the cluster are sufficient, but no TaskManager can
> be allocated. It seems that there is a problem with the calculation logic of
> {{SlotManagerConfiguration#getMaxTotalMem}}. Due to the rounding down of
> division, the calculated {{MemorySize}} is too small.
--
This message was sent by Atlassian Jira
(v8.20.10#820010)
