[jira] [Commented] (HBASE-26190) High rate logging of BucketAllocatorException: Allocation too big

["Duo Zhang (Jira)" <ji...@apache.org>]

    [ https://issues.apache.org/jira/browse/HBASE-26190?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17397753#comment-17397753 ] 

Duo Zhang commented on HBASE-26190:
-----------------------------------

I think this indicates that we misconfigure the bucket cache sizes, so I think WARN is the correct level.

And I do not think we can only log it once, because the probability is the actual problem. If you only saw a few warnings, for example, less than 10 times per day, then usually it is not a problem. But if you see the log flooded with these warnings, then it is a problem.

So I think your latter suggestion is better. Maybe we could count the numbers of bad allocations, and log it once per minute.

Thanks.

> High rate logging of BucketAllocatorException: Allocation too big 
> ------------------------------------------------------------------
>
>                 Key: HBASE-26190
>                 URL: https://issues.apache.org/jira/browse/HBASE-26190
>             Project: HBase
>          Issue Type: Bug
>          Components: BucketCache, Operability
>    Affects Versions: 2.4.5
>            Reporter: Andrew Kyle Purtell
>            Priority: Major
>             Fix For: 2.5.0, 3.0.0-alpha-2, 2.4.6
>
>
> These log lines may be printed at high frequency when using a schema that creates large blocks and the bucket cache is enabled. It makes sense to warn about this initially, but the very high rate of warnings when the use case is legit and this is expected does not. Print this once, then not again; or, rate-limit this message to be printed at a more reasonable rate, like once per minute. 
> {noformat}
> 2021-08-11 23:42:10,902 WARN  [main-BucketCacheWriter-0]
> bucket.BucketCache: Failed allocation for 4842189251414b8a9212f6831462e415_218610843;
> org.apache.hadoop.hbase.io.hfile.bucket.BucketAllocatorException:
> Allocation too big size=1049337;
> adjust BucketCache sizes hbase.bucketcache.bucket.sizes to accomodate
> if size seems reasonable and you want it cached.
> {noformat}
> Also, it might be better to log this at INFO given the caveat "if size seems reasonable and you want it cached". Reads like an informational message.



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