[
https://issues.apache.org/jira/browse/HADOOP-6132?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=12732218#action_12732218
]
Kan Zhang commented on HADOOP-6132:
-----------------------------------
The manual verification I used was running a test that involves the RPC layer
(e.g., a test that uses the MiniDFSCluster), with log level of ipc.Server set
to debug. Without this patch, one would observe the following line in the log,
which means the client is establishing a connection for the VersionedProtocol.
It happens whenever a client sets up the first connection for a protocol (e.g.,
ClientProtocol or DatanodeProtocol). When the bug is fixed, no connection for
the VersionedProtocol should ever be set up. Hence, no such log messages should
have been observed in the log.
{noformat}
2009-07-16 15:41:37,763 DEBUG ipc.Server (Server.java:readAndProcess(861)) -
Successfully authorized org.apache.hadoop.ipc.VersionedProtocol-......
{noformat}
> RPC client opens an extra connection for VersionedProtocol
> ----------------------------------------------------------
>
> Key: HADOOP-6132
> URL: https://issues.apache.org/jira/browse/HADOOP-6132
> Project: Hadoop Common
> Issue Type: Bug
> Components: ipc
> Reporter: Kan Zhang
> Assignee: Kan Zhang
> Attachments: 1.patch, 2.patch
>
>
> RPC client caches connections per protocol. However, since all of our real
> protocols are subclasses of VersionedProtocol, a bug in the implementation
> makes the client opens an extra connection just for the VersionedProtocol,
> which is not needed.
--
This message is automatically generated by JIRA.
-
You can reply to this email to add a comment to the issue online.
