[
https://issues.apache.org/jira/browse/FLINK-23190?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17377343#comment-17377343
]
loyi commented on FLINK-23190:
------------------------------
[~trohrmann] thanks for your explanation. I try implement my idea on
Flink-1.13.1, and it could work well with the default
*declarative-resource-management* feature，but not with AdaptiveScheduler.
Here is the idea:
# When *SlotSharingExecutionSlotAllocator* start a SlotRequestId, we could let
the request bring the "detail" of ExecutionGroup, and then it will be put in
*pendingRequests*.
# Once *NewSlotsListener* triggers, for example provide 4 slots, we could
calculate the proportion of each group by these *requests* , and assign the
slot evenly to these groups.
Considering that we haven't reached consensus yet，i put the concept code on my
reposity:
https://github.com/saddays/flink/commit/9e3c51664f77aee1b7ff8b68726a6b5eaa630e5b
Is this solution feasible? Looking forward to your reply！
> Make task-slot allocation much more evenly
> ------------------------------------------
>
> Key: FLINK-23190
> URL: https://issues.apache.org/jira/browse/FLINK-23190
> Project: Flink
> Issue Type: Improvement
> Components: Runtime / Coordination
> Affects Versions: 1.12.3
> Reporter: loyi
> Priority: Major
>
> FLINK-12122 only guarantees spreading out tasks across the set of TMs which
> are registered at the time of scheduling, but our jobs are all runing on
> active yarn mode， the job with smaller source parallelism offen cause
> load-balance issues.
>
> For this job:
> {code:java}
> // -ys 4 means 10 taskmanagers
> env.addSource(...).name("A").setParallelism(10).
> map(...).name("B").setParallelism(30)
> .map(...).name("C").setParallelism(40)
> .addSink(...).name("D").setParallelism(20);
> {code}
>
> Flink-1.12.3 task allocation:
> ||operator||tm1 ||tm2||tm3||tm4||tm5||5m6||tm7||tm8||tm9||tm10||
> |A|
> 1|{color:#de350b}2{color}|{color:#de350b}2{color}|1|1|{color:#de350b}3{color}|{color:#de350b}0{color}|{color:#de350b}0{color}|{color:#de350b}0{color}|{color:#de350b}0{color}|
> |B|3|3|3|3|3|3|3|3|{color:#de350b}2{color}|{color:#de350b}4{color}|
> |C|4|4|4|4|4|4|4|4|4|4|
> |D|2|2|2|2|2|{color:#de350b}1{color}|{color:#de350b}1{color}|2|2|{color:#de350b}4{color}|
>
> Suggestions:
> When TaskManger start register slots to slotManager , current processing
> logic will choose the first pendingSlot which meet its resource
> requirements. The "random" strategy usually causes uneven task allocation
> when source-operator's parallelism is significantly below process-operator's.
> A simple feasible idea is {color:#de350b}partition{color} the current
> "{color:#de350b}pendingSlots{color}" by their "JobVertexIds" (such as let
> AllocationID bring the detail) , then allocate the slots proportionally to
> each JobVertexGroup.
>
> For above case, the 40 pendingSlots could be divided into 4 groups：
> [ABCD]: 10 // A、B、C、D reprents {color:#de350b}jobVertexId{color}
> [BCD]: 10
> [CD]: 10
> [D]: 10
>
> Every taskmanager will provide 4 slots one time, and each group will get 1
> slot according their proportion (1/4), the final allocation result is below:
> [ABCD] : deploye on 10 different taskmangers
> [BCD]: deploye on 10 different taskmangers
> [CD]: deploye on 10 different taskmangers
> [D]: deploye on 10 different taskmangers
>
> I have implement a [concept
> code|https://github.com/saddays/flink/commit/dc82e60a7c7599fbcb58c14f8e3445bc8d07ace1]
> based on Flink-1.12.3 , the patch version has {color:#de350b}fully
> evenly{color} task allocation , and works well on my workload . Are there
> other point that have not been considered or does it conflict with future
> plans? Sorry for my poor english.
>
>
--
This message was sent by Atlassian Jira
(v8.3.4#803005)
