[
https://issues.apache.org/jira/browse/HADOOP-16452?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16891372#comment-16891372
]
Stephen O'Donnell commented on HADOOP-16452:
--------------------------------------------
The key here is how many blocks are present on *one disk* in the datanode. The
DN sends a report per disk once some threshold is crossed, so in that respect
we already break the reports into pieces. Once the blocks on one disk exceeds
64MB in the report, you start to see this issue.
I have seen a few instances where the 64MB limit has been breached leading to
missing blocks on the namenode. This issue is pretty easy to identify if you
know what to look for, but for the average administrator, it can be quite scary
when you suddenly have a lot of missing blocks.
Additionally, when you get into this sort of situation, the only real solution
is to increase the limit, so it probably makes sense to bump this to 128MB by
default. If the cluster is running with this many blocks, then the NN heap is
probably big enough to accommodate the larger report size anyway.
> Increase ipc.maximum.data.length default from 64MB to 128MB
> -----------------------------------------------------------
>
> Key: HADOOP-16452
> URL: https://issues.apache.org/jira/browse/HADOOP-16452
> Project: Hadoop Common
> Issue Type: Improvement
> Components: ipc
> Affects Versions: 2.6.0
> Reporter: Wei-Chiu Chuang
> Priority: Major
>
> Reason for bumping the default:
> Denser DataNodes are common. It is not uncommon to find a DataNode with > 7
> million blocks these days.
> With such a high number of blocks, the block report message can exceed the
> 64mb limit (defined by ipc.maximum.data.length). The block reports are
> rejected, causing missing blocks in HDFS. We had to double this configuration
> value in order to work around the issue.
> We are seeing an increasing number of these cases. I think it's time to
> revisit some of these default values as the hardware evolves.
--
This message was sent by Atlassian JIRA
(v7.6.14#76016)
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
