[
https://issues.apache.org/jira/browse/HDFS-941?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=12847246#action_12847246
]
Todd Lipcon commented on HDFS-941:
----------------------------------
bq. So at this point, it should be safe to take the Socket away from its
previous owner and give it to a new BlockReader.
I don't think this is true. Two issues:
- At DFSInputStream.java:699, we check the cached socket for usability. If we
have two threads concurrently at this line, they can both detect that it's
usable, and then both move on to construct block readers on the same socket.
One of them will win the entry.setOwner race at line 723, but they'll both
still be operating on the same socket and generate bad commands.
- If a block reader exists and is not complete, usable() returns false so the
socket won't be reused. But on line 701, it calls entry.close() which calls
through to close the underlying socket. This will destroy the read currently
going on.
> Datanode xceiver protocol should allow reuse of a connection
> ------------------------------------------------------------
>
> Key: HDFS-941
> URL: https://issues.apache.org/jira/browse/HDFS-941
> Project: Hadoop HDFS
> Issue Type: Improvement
> Components: data-node, hdfs client
> Affects Versions: 0.22.0
> Reporter: Todd Lipcon
> Assignee: bc Wong
> Attachments: HDFS-941-1.patch
>
>
> Right now each connection into the datanode xceiver only processes one
> operation.
> In the case that an operation leaves the stream in a well-defined state (eg a
> client reads to the end of a block successfully) the same connection could be
> reused for a second operation. This should improve random read performance
> significantly.
--
This message is automatically generated by JIRA.
-
You can reply to this email to add a comment to the issue online.
