[
https://issues.apache.org/jira/browse/HDFS-636?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]
Konstantin Shvachko updated HDFS-636:
-------------------------------------
Attachment: completeBlockTotal.patch
Here is the patch that excludes incomplete blocks from the safe replication
count.
A also reverted changes to TestRenameWhileOpen in one of the previous patches..
The test creates a file and restarts the name-node, then waits until the
cluster is up. Originally it was waiting when the safe-mode is turned off. I
changed it to wait until lease recovery is done. Now I am changing it back to
waiting on exiting from safe-mode, so that it could verify the new safe-mode
condition.
> SafeMode should count only complete blocks.
> -------------------------------------------
>
> Key: HDFS-636
> URL: https://issues.apache.org/jira/browse/HDFS-636
> Project: Hadoop HDFS
> Issue Type: Sub-task
> Components: name-node
> Affects Versions: Append Branch
> Reporter: Konstantin Shvachko
> Assignee: Konstantin Shvachko
> Fix For: Append Branch
>
> Attachments: completeBlockTotal.patch
>
>
> During start up the name-node is in safe mode and is counting blocks reported
> by data-nodes. When the number of minimally replicated blocks reaches the
> configured threshold the name-node leaves safe mode. Currently all blocks are
> counted towards the threshold including the ones that are under construction.
> The under-construction blocks should be excluded from the count, because they
> need to be recovered, which may take long time (lease expires in 1 hour by
> default). Also the recovery may result in deleting those blocks so counting
> them in the blocks total is incorrect.
--
This message is automatically generated by JIRA.
-
You can reply to this email to add a comment to the issue online.
