[
https://issues.apache.org/jira/browse/FLINK-27155?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17520188#comment-17520188
]
Yuan Mei edited comment on FLINK-27155 at 4/10/22 3:29 PM:
-----------------------------------------------------------
Hey [~Feifan Wang] , by "{+}Indeed, downloading the same files multiple times
does not sound necessary.{+}", I mean I agree with you that it does not
necessary to download the same files multiple times.
By "Task Local Recovery", I mean it can partially resolve the problem you
mentioned, and let's focus on that first.
By caching the same file, I think it is a good improvement, but may share some
common paths from task local recovery, for example, "life cycle management"
(who is responsible for download/clean-up/successive failures, e.t.c).
Hope this clarifies my previous replies.
was (Author: ym):
Hey [~Feifan Wang] , by "{+}Indeed, downloading the same files multiple times
does not sound necessary.{+}", I mean I agree with you that it does not
necessary to download the same files multiple times.
By "Task Local Recovery", I mean it can partially resolve the problem you
mentioned, and let's focus on that first.
By caching the same file, I think it is a good improvement, but may share some
common paths from task local recovery, for example, "life cycle management"
(who is responsible for download/clean-up/successive failures, e.t.c), and this
part may be related to TM-Ownership that Roman is working on right now as well.
Hope this clarifies my previous replies.
> Reduce multiple reads to the same Changelog file in the same taskmanager
> during restore
> ---------------------------------------------------------------------------------------
>
> Key: FLINK-27155
> URL: https://issues.apache.org/jira/browse/FLINK-27155
> Project: Flink
> Issue Type: Improvement
> Components: Runtime / Checkpointing, Runtime / State Backends
> Reporter: Feifan Wang
> Priority: Major
>
> h3. Background
> In the current implementation, State changes of different operators in the
> same taskmanager may be written to the same changelog file, which effectively
> reduces the number of files and requests to DFS.
> But on the other hand, the current implementation also reads the same
> changelog file multiple times on recovery. More specifically, the number of
> times the same changelog file is accessed is related to the number of
> ChangeSets contained in it. And since each read needs to skip the preceding
> bytes, this network traffic is also wasted.
> The result is a lot of unnecessary request to DFS when there are multiple
> slots and keyed state in the same taskmanager.
> h3. Proposal
> We can reduce multiple reads to the same changelog file in the same
> taskmanager during restore.
> One possible approach is to read the changelog file all at once and cache it
> in memory or local file for a period of time when reading the changelog file.
> I think this could be a subtask of [v2 FLIP-158: Generalized incremental
> checkpoints|https://issues.apache.org/jira/browse/FLINK-25842] .
> Hi [~ym] , [~roman] how do you think about ?
--
This message was sent by Atlassian Jira
(v8.20.1#820001)
