[
https://issues.apache.org/jira/browse/HDFS-7342?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=14225767#comment-14225767
]
Vinayakumar B commented on HDFS-7342:
-------------------------------------
bq. It seems that lease recovery could happen before the client request comes
here, when this happens, the block state would be COMMITTED with minReplication
met, right?
We are talking about the state of the penultimate block not the last block,
which is the cause found for this issue.
1. For the penultimate block, only client request (request for another block)
will make it COMMITTED, as client will be still alive and adds one more block.
2. And for the last block, client makes it COMMITTED during normal closure,
else {{commitBlockSynchronization()}} during the lease recovery closure.
I see no other places, block getting COMMITTED.
> Lease Recovery doesn't happen some times
> ----------------------------------------
>
> Key: HDFS-7342
> URL: https://issues.apache.org/jira/browse/HDFS-7342
> Project: Hadoop HDFS
> Issue Type: Bug
> Affects Versions: 2.0.0-alpha
> Reporter: Ravi Prakash
> Assignee: Ravi Prakash
> Attachments: HDFS-7342.1.patch, HDFS-7342.2.patch, HDFS-7342.3.patch
>
>
> In some cases, LeaseManager tries to recover a lease, but is not able to.
> HDFS-4882 describes a possibility of that. We should fix this
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
