[
https://issues.apache.org/jira/browse/HDFS-5029?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]
Daryn Sharp updated HDFS-5029:
------------------------------
Attachment: HDFS-5029.patch
HDFS-5029.branch-23.patch
Token operations should ideally not lock the namespace at all, but the read
lock is at least required to prevent the the NN from transitioning states
during an operation.
Changed writeLock to readLock for token operations. Edit logging appears to be
thread safe regardless of lock held.
Use synch on dtsm to aggregate renew/cancel + edit log to prevent race
condition where a near concurrent renew followed by cancel might be logged out
of order. This unfortunate synch is "ok" because the secret manager is already
heavily synch'ed so I made it no worse than it already is.
> Token operations should not block write operations
> --------------------------------------------------
>
> Key: HDFS-5029
> URL: https://issues.apache.org/jira/browse/HDFS-5029
> Project: Hadoop HDFS
> Issue Type: Improvement
> Affects Versions: 0.23.0, 2.0.0-alpha, 3.0.0
> Reporter: Daryn Sharp
> Assignee: Daryn Sharp
> Attachments: HDFS-5029.branch-23.patch, HDFS-5029.patch
>
>
> Token operations unnecessarily obtain the write lock on the namespace. Edits
> for token operations are independent of edits for other namespace write
> operations, and the edits have no ordering requirement with respect to
> namespace changes.
--
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
