hadoop-hdfs-issues mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From "Daryn Sharp (JIRA)" <j...@apache.org>
Subject [jira] [Commented] (HDFS-6763) Standby NN stalls after processing edits
Date Tue, 29 Jul 2014 16:17:39 GMT

    [ https://issues.apache.org/jira/browse/HDFS-6763?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=14077908#comment-14077908
] 

Daryn Sharp commented on HDFS-6763:
-----------------------------------

It appears to be an attempt to self-correct troublesome quota bugs that in the past have crashed
secondaries.  I seem to recall secondaries crashing were odd:  you can read an image, always
apply edits, write the image, but further attempts to apply edits that violated quotas would
crash.  The "fix" was always create a new namesystem for the next checkpoint but I don't think
that's feasible for a standby.

I don't like masking bugs so maybe it should never be invoked.  Unless quotas aren't really
being computed during edit tailing/replay, in which case updating the quota during a transition
to active is more appropriate.

> Standby NN stalls after processing edits
> ----------------------------------------
>
>                 Key: HDFS-6763
>                 URL: https://issues.apache.org/jira/browse/HDFS-6763
>             Project: Hadoop HDFS
>          Issue Type: Bug
>          Components: ha, namenode
>            Reporter: Daryn Sharp
>
> {{FSImage#loadEdits}} calls {{updateCountForQuota}} to recalculate & verify quotas
for the entire namespace.  A standby NN using shared edits calls this method every minute.
 The standby may appear to "hang" for many seconds.



--
This message was sent by Atlassian JIRA
(v6.2#6252)

Mime
View raw message