[
https://issues.apache.org/jira/browse/HDFS-17536?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17849244#comment-17849244
]
ASF GitHub Bot commented on HDFS-17536:
---------------------------------------
ZanderXu opened a new pull request, #6844:
URL: https://github.com/apache/hadoop/pull/6844
Both `RouterAdminServer#enterSafeMode()` and
`RouterSafemodeService#periodicInvoke()#leave` can change the router state at
the same time.
Safe-mode change logic should be condensed into one method. And some races
may happen in the current implementation, such as:
1. `RouterAdminServer#enterSafeMode()` set router stat to
`RouterServiceState.SAFEMODE`
2. `RouterSafemodeService#periodicInvoke()#leave` got true when checking
`safeMode && !isSafeModeSetManually`
3. `RouterAdminServer#enterSafeMode()` set `safeMode` and
`isSafeModeSetManually` to `true`
4. `RouterAdminServer#enterSafeMode()` get `true` when checking safe-mode
5. `RouterSafemodeService#periodicInvoke()#leave` call `leave()` to leave
safe-mode.
This RBF is not in safe-mode and `safeMode` is `false`, but
`isSafeModeSetManually` is `true`.
> RBF: Format safe-mode related logic and fix a race
> ---------------------------------------------------
>
> Key: HDFS-17536
> URL: https://issues.apache.org/jira/browse/HDFS-17536
> Project: Hadoop HDFS
> Issue Type: Task
> Reporter: ZanderXu
> Assignee: ZanderXu
> Priority: Major
>
> RBF: Format safe-mode related logic and fix a race.
--
This message was sent by Atlassian Jira
(v8.20.10#820010)
---------------------------------------------------------------------
To unsubscribe, e-mail: hdfs-issues-unsubscr...@hadoop.apache.org
For additional commands, e-mail: hdfs-issues-h...@hadoop.apache.org
