[
https://issues.apache.org/jira/browse/HDFS-15650?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17224923#comment-17224923
]
Yushi Hayasaka commented on HDFS-15650:
---------------------------------------
[~weichiu] Thanks! I see... :sad:
Then, Just idea, how about using the new configuration, which is named
`dfs.checksum.ec.socket-timeout` (or `dfs.ec.checksum.socket-timeout`?). It
will be added in hdfs-default.xml like below:
{noformat}
<property>
<name>dfs.checksum.ec.socket-timeout</name>
<value>3000</value>
<description>
Default timeout value in milliseconds for computing the checksum of striped
blocks.
Recommended to set the same value between client and DNs in an erasure
coded cluster because mismatching may cause exhausting handler threads.
</description>
</property>{noformat}
I think it is easier for administrators to notice the risk because they needs
to see here for understanding the configuration, and the confusion which is
caused by changing behavior may be depressed since the default value is the
same with the current value.
> Make the socket timeout for computing checksum of striped blocks configurable
> -----------------------------------------------------------------------------
>
> Key: HDFS-15650
> URL: https://issues.apache.org/jira/browse/HDFS-15650
> Project: Hadoop HDFS
> Issue Type: Improvement
> Components: datanode, ec, erasure-coding
> Reporter: Yushi Hayasaka
> Priority: Minor
> Labels: pull-request-available
> Time Spent: 20m
> Remaining Estimate: 0h
>
> Regarding the DataNode tries to get the checksum of EC internal blocks from
> another DataNode for computing the checksum of striped blocks, the timeout is
> hard-coded now, but it should be configurable.
--
This message was sent by Atlassian Jira
(v8.3.4#803005)
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
