[jira] [Comment Edited] (FLINK-18662) Provide more detailed metrics why unaligned checkpoint is taking long time

["Piotr Nowojski (Jira)" <ji...@apache.org>]

    [ https://issues.apache.org/jira/browse/FLINK-18662?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17164495#comment-17164495 ] 

Piotr Nowojski edited comment on FLINK-18662 at 7/24/20, 3:34 PM:
------------------------------------------------------------------

I'm not sure which option is better:

# Add new metric {{checkpointUnalignmentTime}}, which would be mutually exclusive with {{checkpointAlignmentTime}} - if one is non zero, the other must be zero and vice versa. This might be a bit annoying to for someone setting up some metric reporting system - he would have to setup and monitor both things.
# Overload the existing {{checkpointAlignmentTime}} metric, and adapt it to repot unalignment time for unaligned checkpoints. Note that in the webUI user would need to understand with what type of checkpoint he is dealing with and what's the difference. Also it might be a bit confusing if web UI presents an aligned savepoint and unaligned checkpoint at the same time. On the other hand the delay between the first and last checkpoint barrier for unaligned checkpoints could arguably be called "alignment time".

Regarding the size of unaligned data, we need to add a new metric, as {{lastCheckpointUnalignmentSize}}/"Alignment Buffered" was removed in FLINK-16404.

[~zjwang], [~AHeise], [~roman_khachatryan] what do you think?


was (Author: pnowojski):
I'm not sure which option is better:

# Add new metric {{checkpointUnalignmentTime}}, which would be mutually exclusive with {{checkpointAlignmentTime}} - if one is non zero, the other must be zero and vice versa. This might be a bit annoying to for someone setting up some metric reporting system - he would have to setup both things.
# Overload the existing {{checkpointAlignmentTime}} metric, and adapt it to repot unalignment time for unaligned checkpoints. Note that in the webUI user would need to understand with what type of checkpoint he is dealing with and what's the difference. Also it might be a bit confusing if web UI presents an aligned savepoint and unaligned checkpoint at the same time. On the other hand the delay between the first and last checkpoint barrier for unaligned checkpoints could arguably be called "alignment time".

Regarding the size of unaligned data, we need to add a new metric, as {{lastCheckpointUnalignmentSize}}/"Alignment Buffered" was removed in FLINK-16404.

[~zjwang], [~AHeise], [~roman_khachatryan] what do you think?

> Provide more detailed metrics why unaligned checkpoint is taking long time
> --------------------------------------------------------------------------
>
>                 Key: FLINK-18662
>                 URL: https://issues.apache.org/jira/browse/FLINK-18662
>             Project: Flink
>          Issue Type: Improvement
>          Components: Runtime / Metrics, Runtime / Network
>    Affects Versions: 1.11.1
>            Reporter: Piotr Nowojski
>            Priority: Critical
>             Fix For: 1.12.0
>
>         Attachments: Screenshot 2020-07-21 at 11.50.02.png
>
>
> With unaligned checkpoint there can happen situation as in the attached screenshot.
>  Task reports long end to end checkpoint time (~2h50min), ~0s sync time, ~2h50min async time, ~0s start delay. It means that task received first checkpoint barrier from one of the channels very quickly (~0s), sync part was quick, but we do not know why async part was taking so long. It could be because of three things:
> # long operator state IO writes
> # long spilling of in-flight data
> # long time to receive the final checkpoint barrier from the last lagging channel
> First and second are probably indistinguishable and the difference between them doesn't matter much for analyzing. However the last one is quite different. It might be independent of the IO, and we are missing this information. 
> Maybe we could report it as "alignment duration" and while we are at it, we could also report amount of spilled in-flight data for unaligned checkpoints as "alignment buffered"? 
> Ideally we should report it as new metrics, but that leaves a question how to display it in the UI, with limited space available. Maybe it could be reported as:
> ||Alignment Buffered||Alignment Duration||
> |0 B (632 MB)|0ms (2h 49m 32s)|
> Where the values in the parenthesis would come from unaligned checkpoints. 



--
This message was sent by Atlassian Jira
(v8.3.4#803005)