[jira] [Assigned] (SAMZA-772) Kafka in-memory buffer may cause OOM: Heap Space errors

["Yi Pan (Data Infrastructure) (JIRA)" <ji...@apache.org>]

     [ https://issues.apache.org/jira/browse/SAMZA-772?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel ]

Yi Pan (Data Infrastructure) reassigned SAMZA-772:
--------------------------------------------------

    Assignee: Yi Pan (Data Infrastructure)

> Kafka in-memory buffer may cause OOM: Heap Space errors
> -------------------------------------------------------
>
>                 Key: SAMZA-772
>                 URL: https://issues.apache.org/jira/browse/SAMZA-772
>             Project: Samza
>          Issue Type: Bug
>          Components: kafka
>    Affects Versions: 0.9.1
>            Reporter: Nick Lee
>            Assignee: Yi Pan (Data Infrastructure)
>             Fix For: 0.10.0
>
>
> The default setting for systems.system-name.samza.fetch.threshold will pre-fetch 50,000 messages from Kafka, ignoring the availability of heap space. When working with relatively large messages, this can lead to Samza job failures. 
> A simple solution would be to restrict the maximum size of org.apache.samza.system.SystemConsumers.unprocessedMessagesBySSP.table to no more than a configurable percentage (35% by default?) of the maximum heap space. 



--
This message was sent by Atlassian JIRA
(v6.3.4#6332)