[ 
https://issues.apache.org/jira/browse/HDFS-3862?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=14140184#comment-14140184
 ] 

Uma Maheswara Rao G commented on HDFS-3862:
-------------------------------------------

Thanks a lot Yi for addressing the feedback.
Me and Vinay has chat on this offline. The conclusion is same like to address 
above concern.
What we thought is, we can control this fencer requirement via configuration 
item at ZKFC instead of having changes in JMs impls. Ideally concern is same 
that ZKFC need not initialize JM impls. There may be chances that JM impls can 
do some initialization work at that time. Example BKJM does some znode 
creations etc I remember.  That things getting created from ZKFC is quite not 
right.
So, How about we keep simple config at ZKFC itself and we avoid implementing 
isNativelySingleWriter API?
Rest all conditions can be same how you handled in this patch. With command 
like force fence option, we try to use configured fencer and do fencing. In 
Normal failover, we do fencing depending on this new configuration item even 
though fencers configured.


> QJM: don't require a fencer to be configured if shared storage has built-in 
> single-writer semantics
> ---------------------------------------------------------------------------------------------------
>
>                 Key: HDFS-3862
>                 URL: https://issues.apache.org/jira/browse/HDFS-3862
>             Project: Hadoop HDFS
>          Issue Type: Sub-task
>          Components: ha
>    Affects Versions: QuorumJournalManager (HDFS-3077)
>            Reporter: Todd Lipcon
>            Assignee: Yi Liu
>         Attachments: HDFS-3862.001.patch, HDFS-3862.002.patch, 
> HDFS-3862.003.patch
>
>
> Currently, NN HA requires that the administrator configure a fencing method 
> to ensure that only a single NameNode may write to the shared storage at a 
> time. Some shared edits storage implementations (like QJM) inherently enforce 
> single-writer semantics at the storage level, and thus the user should not be 
> forced to specify one.
> We should extend the JournalManager interface so that the HA code can operate 
> without a configured fencer if the JM has such built-in fencing.



--
This message was sent by Atlassian JIRA
(v6.3.4#6332)

Reply via email to