[
https://issues.apache.org/jira/browse/YARN-1776?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]
Zhijie Shen updated YARN-1776:
------------------------------
Attachment: YARN-1776.2.patch
I created a new patch, though I cannot do much to make sure updating DT is
completely atomic (which touches more than one file) in FileSystemRMStateStore,
I try to make sure update a file behave somehow atomically. It is useful for
update DT, App and AppAttempt.
To update a file:
1. write new file to .new
2. delete the current file
3. rename .new to the current file
Here, write is FileSystemRMStateStore#writeFile to ensure atomic. Then, if RM
fails before 1, nothing needs to be changed. If RM fails between 1 and 2, and
between 2 and 3, we have the .new in hand, we can replace the current file with
it when loading the state information. If RM fails after 3, nothing needs to be
changed as well.
Please let me know if you want to combine these fix together, or split it into
two tickets.
> renewDelegationToken should survive RM failover
> -----------------------------------------------
>
> Key: YARN-1776
> URL: https://issues.apache.org/jira/browse/YARN-1776
> Project: Hadoop YARN
> Issue Type: Sub-task
> Reporter: Zhijie Shen
> Assignee: Zhijie Shen
> Attachments: YARN-1776.1.patch, YARN-1776.2.patch
>
>
> When a delegation token is renewed, two RMStateStore operations: 1) removing
> the old DT, and 2) storing the new DT will happen. If RM fails in between.
> There would be problem.
--
This message was sent by Atlassian JIRA
(v6.2#6252)
