[
https://issues.apache.org/jira/browse/HDFS-5025?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]
Jing Zhao updated HDFS-5025:
----------------------------
Attachment: HDFS-5025.001.patch
Update the patch:
1. Add the functionality to rebuild retry cache based on editlog.
2. Some editlog ops (e.g., UpdateBlocksOp) may be associated with multiple
protocol calls which are not all "AtMostOnce"/idempotent. Currently I add a
boolean parameter in FSEditLog#logXXX methods to indicate whether we need to
record RPC ids in editlog.
The patch depends on HDFS-4979 and is only for initial review. Next step is to
add test code including HA-related tests.
> Record ClientId and CallId in EditLog to enable rebuilding retry cache in
> case of HA failover
> ---------------------------------------------------------------------------------------------
>
> Key: HDFS-5025
> URL: https://issues.apache.org/jira/browse/HDFS-5025
> Project: Hadoop HDFS
> Issue Type: Sub-task
> Components: ha, namenode
> Affects Versions: 3.0.0
> Reporter: Jing Zhao
> Assignee: Jing Zhao
> Attachments: HDFS-5025.000.patch, HDFS-5025.001.patch
>
>
> In case of HA failover, we need to be able to rebuild the retry cache in the
> other Namenode. We thus need to record client id and call id in the edit log
> for those "AtMostOnce" operations.
--
This message is automatically generated by JIRA.
If you think it was sent incorrectly, please contact your JIRA administrators
For more information on JIRA, see: http://www.atlassian.com/software/jira
