[
https://issues.apache.org/jira/browse/HADOOP-1707?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel#action_12534952
]
Doug Cutting commented on HADOOP-1707:
--------------------------------------
> The client gets an exception if the primary datanode fails.
Why can't it simply replace the primary with one of the secondary datanodes and
proceed?
> If a secondary datanode fails, the primary informs the client about this
> event.
Since a secondary will typically fail by timing out, the timeout used between
the client and the primary must be longer than that used between the primary
and secondary, so that the client waits long enough to hear about a failed
secondary. And the timeout used between the application and the client must be
longer yet. Right? Perhaps we should make all these timeouts proportional to
a single configuration parameter, the application timeout?
If we wish to ensure that blocks are sufficiently replicated, then we'll block
on file close, right?
Overall, this sounds like an approach worth trying.
> Remove the DFS Client disk-based cache
> --------------------------------------
>
> Key: HADOOP-1707
> URL: https://issues.apache.org/jira/browse/HADOOP-1707
> Project: Hadoop
> Issue Type: Improvement
> Components: dfs
> Reporter: dhruba borthakur
> Assignee: dhruba borthakur
> Fix For: 0.16.0
>
>
> The DFS client currently uses a staging file on local disk to cache all
> user-writes to a file. When the staging file accumulates 1 block worth of
> data, its contents are flushed to a HDFS datanode. These operations occur
> sequentially.
> A simple optimization of allowing the user to write to another staging file
> while simultaneously uploading the contents of the first staging file to HDFS
> will improve file-upload performance.
--
This message is automatically generated by JIRA.
-
You can reply to this email to add a comment to the issue online.
