[
https://issues.apache.org/jira/browse/HDFS-13475?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16541965#comment-16541965
]
Íñigo Goiri commented on HDFS-13475:
------------------------------------
I think the proposal in [^HDFS-13475.002.patch] is much cleaner; isolating the
safe mode logic makes the code simpler.
I think the unit test covers the issue described in this JIRA.
My only concern is how long it takes as it is adding an extra 7 seconds which
is basically doubling the test length.
I wonder if there is any way to speed this up a little; for example, cutting
the times a little and waiting for it to get into RUNNING instead of just
sleeping.
A minor nit regarding code, we currently have:
{code}
RouterSafemodeService safeModeService = this.router.getSafemodeService();
if (safeModeService != null) {
this.router.updateRouterState(RouterServiceState.SAFEMODE);
safeModeService.setManualSafeMode(true);
return EnterSafeModeResponse.newInstance(verifySafeMode(true));
}
return EnterSafeModeResponse.newInstance(false);
{code}
It is personal preference, but I would go for a single point of exit:
{code}
boolean success = false;
RouterSafemodeService safeModeService = this.router.getSafemodeService();
if (safeModeService != null) {
this.router.updateRouterState(RouterServiceState.SAFEMODE);
safeModeService.setManualSafeMode(true);
success = verifySafeMode(true);
}
return EnterSafeModeResponse.newInstance(success);
{code}
Similar for the other modified methods in {{RouterAdminServer}}.
> 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,
> HDFS-13475.002.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
