[jira] [Updated] (SAMZA-64) Fail KafkaCheckpointManager on unrecoverable errors

["Yan Fang (JIRA)" <ji...@apache.org>]

     [ https://issues.apache.org/jira/browse/SAMZA-64?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel ]

Yan Fang updated SAMZA-64:
--------------------------

    Attachment: SAMZA-64.patch

> Fail KafkaCheckpointManager on unrecoverable errors
> ---------------------------------------------------
>
>                 Key: SAMZA-64
>                 URL: https://issues.apache.org/jira/browse/SAMZA-64
>             Project: Samza
>          Issue Type: Bug
>          Components: kafka
>    Affects Versions: 0.6.0
>            Reporter: Chris Riccomini
>         Attachments: SAMZA-64.patch
>
>
> KafkaCheckpointManager currently handles all errors from Kafka by waiting, and retrying. These error codes are located in Kafka's ErrorMapping class:
> {noformat}
> val UnknownCode : Short = -1
> val NoError : Short = 0
> val OffsetOutOfRangeCode : Short = 1
> val InvalidMessageCode : Short = 2
> val UnknownTopicOrPartitionCode : Short = 3
> val InvalidFetchSizeCode  : Short = 4
> val LeaderNotAvailableCode : Short = 5
> val NotLeaderForPartitionCode : Short = 6
> val RequestTimedOutCode: Short = 7
> val BrokerNotAvailableCode: Short = 8
> val ReplicaNotAvailableCode: Short = 9
> val MessageSizeTooLargeCode: Short = 10
> val StaleControllerEpochCode: Short = 11
> {noformat}
> There was some discussion on SAMZA-62's RB about potentially not retrying on ALL errors. Some are not recoverable (e.g. InvalidMessageCode).



--
This message was sent by Atlassian JIRA
(v6.2#6252)