[
https://issues.apache.org/jira/browse/HDFS-13475?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16540682#comment-16540682
]
Chao Sun commented on HDFS-13475:
---------------------------------
Thanks [~elgoiri]. I think we still need two flags still: one {{safeMode}} and
another {{isSafeModeSetManually}}, but we can set them together in
{{setManualSafeMode(true)}} and remove {{setSafeMode()}}.
One question related to the test failures though: after we move the logic to
{{RouterSafemodeService}}, enter/leave/get safe mode will no longer be
effective *if* the safe mode service is disabled. We also need to check several
places on the result of {{getSafemodeService}} before we call any method on the
service. Let me know if you are OK with this.
> RBF: Admin cannot enforce Router enter SafeMode
> -----------------------------------------------
>
> Key: HDFS-13475
> URL: https://issues.apache.org/jira/browse/HDFS-13475
> Project: Hadoop HDFS
> Issue Type: Sub-task
> Reporter: Wei Yan
> Assignee: Chao Sun
> Priority: Major
> Attachments: HDFS-13475.000.patch, HDFS-13475.001.patch
>
>
> To reproduce the issue:
> {code:java}
> $ bin/hdfs dfsrouteradmin -safemode enter
> Successfully enter safe mode.
> $ bin/hdfs dfsrouteradmin -safemode get
> Safe Mode: true{code}
> And then,
> {code:java}
> $ bin/hdfs dfsrouteradmin -safemode get
> Safe Mode: false{code}
> From the code, it looks like the periodicInvoke triggers the leave.
> {code:java}
> public void periodicInvoke() {
> ......
> // Always update to indicate our cache was updated
> if (isCacheStale) {
> if (!rpcServer.isInSafeMode()) {
> enter();
> }
> } else if (rpcServer.isInSafeMode()) {
> // Cache recently updated, leave safe mode
> leave();
> }
> }
> {code}
>
>
--
This message was sent by Atlassian JIRA
(v7.6.3#76005)
---------------------------------------------------------------------
To unsubscribe, e-mail: hdfs-issues-unsubscr...@hadoop.apache.org
For additional commands, e-mail: hdfs-issues-h...@hadoop.apache.org
