You are viewing a plain text version of this content. The canonical link for it is here.
Posted to jira@kafka.apache.org by "John Roesler (Jira)" <ji...@apache.org> on 2020/06/17 14:38:00 UTC

[jira] [Resolved] (KAFKA-10165) Percentiles metric leaking memory

     [ https://issues.apache.org/jira/browse/KAFKA-10165?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel ]

John Roesler resolved KAFKA-10165.
----------------------------------
    Resolution: Fixed

> Percentiles metric leaking memory
> ---------------------------------
>
>                 Key: KAFKA-10165
>                 URL: https://issues.apache.org/jira/browse/KAFKA-10165
>             Project: Kafka
>          Issue Type: Bug
>          Components: metrics, streams
>            Reporter: Sophie Blee-Goldman
>            Assignee: John Roesler
>            Priority: Blocker
>             Fix For: 2.6.0
>
>
> We've hit several OOM in our soak cluster lately. We were finally able to get a heap dump right after the OOM, and found over 3.5 GB of memory being retained by the percentiles (or specifically by the 1MB float[] used by the percentiles). 
> The leak does seem specific to the Percentiles class, as we see ~3000 instances of the Percentiles object vs only ~500 instances of the Max object, which is also used in the same sensor as the Percentiles
> We did recently lower the size from 1MB to 100kB, but it's clear there is a leak of some kind and a "smaller leak" is not an acceptable solution. If the cause fo the leak is not immediately obvious we should just revert the percentiles in 2.6 and work on stabilizing them for 2.7



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