[
https://issues.apache.org/jira/browse/HADOOP-1126?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]
Hairong Kuang updated HADOOP-1126:
----------------------------------
Attachment: neededReplicationRemove.patch
This patch fixed the described problem. It also changed the name of the class
by incorporating the suggestion in HADOOP-1026.
I tested the patch at a one-node cluster and saw 0 block in the
underReplicatedBlocks when the namenode restarted and left the safe mode.
> Optimize CPU usage when cluster restarts
> ----------------------------------------
>
> Key: HADOOP-1126
> URL: https://issues.apache.org/jira/browse/HADOOP-1126
> Project: Hadoop
> Issue Type: Bug
> Components: dfs
> Reporter: dhruba borthakur
> Assigned To: Hairong Kuang
> Attachments: neededReplicationRemove.patch
>
>
> When the namenode restarts, all datanodes report their blocks to the
> namenode. The namenode inserts these blocks into neededReplication(). When
> safemode exists, pendingTransfers removes them from neededReplication. This
> is non-optimal because pendingTransfer consumes plenty of CPU just after
> exiting safe-mode.
> The problem is that neededReplications.remove(block) does not remove a block
> if oldExpectedReplicas is exactly equal to oldReplicas.
--
This message is automatically generated by JIRA.
-
You can reply to this email to add a comment to the issue online.
