[
https://issues.apache.org/jira/browse/HDFS-395?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13065523#comment-13065523
]
Suresh Srinivas commented on HDFS-395:
--------------------------------------
>From my understanding, when a namenode removes a block:
# NN first deletes it in the block map and adds it to invalidate set for the
datanodes.
# The blocks are then deleted from invalidate set, when it is sent for deletion
to the datanodes.
Namenode already has deleted from it's data structures the block information.
Given this I am not sure what purpose the delete ack serves.
I wanted to do a variant of this for the following reasons:
If accidentally files related to block replicas are deleted on the datanode,
without a periodic block reports the namenode is not aware of this loss. Also
a replica file could be truncated/modified. The change I was thinking was in
DirectoryScanner; it currently reconciles the difference between the block
information in datanode process with what is on the disk. This difference is
sent to namenode in the block report. Instead of block report, we could send
this diff to the namenode. With this we can reduce the frequency of block
report.
+1 for renaming the file to target it for deletion. This addresses some of the
race conditions that Hairong brought up.
> DFS Scalability: Incremental block reports
> ------------------------------------------
>
> Key: HDFS-395
> URL: https://issues.apache.org/jira/browse/HDFS-395
> Project: Hadoop HDFS
> Issue Type: Sub-task
> Reporter: dhruba borthakur
> Assignee: dhruba borthakur
> Attachments: blockReportPeriod.patch, explicitDeleteAcks.patch
>
>
> I have a cluster that has 1800 datanodes. Each datanode has around 50000
> blocks and sends a block report to the namenode once every hour. This means
> that the namenode processes a block report once every 2 seconds. Each block
> report contains all blocks that the datanode currently hosts. This makes the
> namenode compare a huge number of blocks that practically remains the same
> between two consecutive reports. This wastes CPU on the namenode.
> The problem becomes worse when the number of datanodes increases.
> One proposal is to make succeeding block reports (after a successful send of
> a full block report) be incremental. This will make the namenode process only
> those blocks that were added/deleted in the last period.
--
This message is automatically generated by JIRA.
For more information on JIRA, see: http://www.atlassian.com/software/jira
