[
https://issues.apache.org/jira/browse/FLINK-9061?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16423072#comment-16423072
]
Jamie Grier edited comment on FLINK-9061 at 4/2/18 8:17 PM:
------------------------------------------------------------
So, what I'm suggesting is that we, optionally, split on '/' and reverse the
components like so:
s3://my_bucket/flink/checkpoints/JOB_ID/chk_000/123456789
becomes
s3://my_bucket/123456789/CHK_000/JOB_ID/checkpoints
It's not very ops friendly but that's because S3 isn't a filesystem. The
hierarchy isn't real. It's a flat keyspace (I know we all know that of
course). The equivalent of a directory listing that would group all the
checkpoints for a single job would be:
aws s3 list s3://my_bucket | grep "JOB_ID/checkpoints"
I think that would work just fine for our use case. What about others?
was (Author: jgrier):
So, what I'm suggesting is that we, optionally, split on '/' and reverse the
components like so:
s3://my_bucket/flink/checkpoints/JOB_ID/chk_000/123456789
becomes
s3://my_bucket/123456789/CHK_000/JOB_ID/checkpoints
It's not very ops friendly but that's because S3 isn't a filesystem. The
hierarchy isn't real. The equivalent of a directory listing that would group
all the checkpoints for a single job would be:
aws s3 list s3://my_bucket | grep "JOB_ID/checkpoints"
I think that would work just fine for our use case. What about others?
> S3 checkpoint data not partitioned well -- causes errors and poor performance
> -----------------------------------------------------------------------------
>
> Key: FLINK-9061
> URL: https://issues.apache.org/jira/browse/FLINK-9061
> Project: Flink
> Issue Type: Bug
> Components: FileSystem, State Backends, Checkpointing
> Affects Versions: 1.4.2
> Reporter: Jamie Grier
> Priority: Critical
>
> I think we need to modify the way we write checkpoints to S3 for high-scale
> jobs (those with many total tasks). The issue is that we are writing all the
> checkpoint data under a common key prefix. This is the worst case scenario
> for S3 performance since the key is used as a partition key.
>
> In the worst case checkpoints fail with a 500 status code coming back from S3
> and an internal error type of TooBusyException.
>
> One possible solution would be to add a hook in the Flink filesystem code
> that allows me to "rewrite" paths. For example say I have the checkpoint
> directory set to:
>
> s3://bucket/flink/checkpoints
>
> I would hook that and rewrite that path to:
>
> s3://bucket/[HASH]/flink/checkpoints, where HASH is the hash of the original
> path
>
> This would distribute the checkpoint write load around the S3 cluster evenly.
>
> For reference:
> https://aws.amazon.com/premiumsupport/knowledge-center/s3-bucket-performance-improve/
>
> Any other people hit this issue? Any other ideas for solutions? This is a
> pretty serious problem for people trying to checkpoint to S3.
>
> -Jamie
>
--
This message was sent by Atlassian JIRA
(v7.6.3#76005)
