[
https://issues.apache.org/jira/browse/HADOOP-2188?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]
Hairong Kuang updated HADOOP-2188:
----------------------------------
Release Note:
This patch introduces pings when client reads are timed out. Server does not
throw away old RPCs. So if clients and the server are on different versions,
they are not able to function well.
The property ipc.client.timeout is removed from the default hadoop
configuration. It also removes metrics RpcOpsDiscardedOPsNum.
Hadoop Flags: [Incompatible change, Reviewed] (was: [Reviewed])
Status: Patch Available (was: Open)
> RPC should send a ping rather than use client timeouts
> ------------------------------------------------------
>
> Key: HADOOP-2188
> URL: https://issues.apache.org/jira/browse/HADOOP-2188
> Project: Hadoop Core
> Issue Type: Improvement
> Components: dfs, ipc
> Affects Versions: 0.16.1
> Reporter: Owen O'Malley
> Assignee: Hairong Kuang
> Fix For: 0.18.0
>
> Attachments: ipc-timeout.patch, ipc-timeout1.patch,
> ipc-timeout2.patch, ipc-timeout3.patch, ipc-timeout4.patch,
> ipc-timeout5.patch, ipc-timeout6.patch, ipc-timeout7.patch,
> ipc-timeout8.patch, ipc-timeout9.patch, rpc-to.patch
>
>
> Current RPC (really IPC) relies on client side timeouts to find "dead"
> sockets. I propose that we have a thread that once a minute (if the
> connection has been idle) writes a "ping" message to the socket. The client
> can detect a dead socket by the resulting error on the write, so no client
> side timeout is required. Also note that the ipc server does not need to
> respond to the ping, just discard it.
--
This message is automatically generated by JIRA.
-
You can reply to this email to add a comment to the issue online.
