[
https://issues.apache.org/jira/browse/HADOOP-13465?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]
Daryn Sharp updated HADOOP-13465:
---------------------------------
Attachment: HADOOP-13465.1.patch
It's not a serious issue. When I added postponed RPC calls, I had to add to
accommodate tests that expected ExitUtils exceptions to bubble back to the
client. So I added the ability to abort responses and did, um, "clever" things
to propagate back a can't-happen-in-reality exception.
To avoid races, the original call is cloned, updated with the fatal exception,
sent. The test failed because I sent the original call's response. This
patch changes {{Connection#doResponse}} from {{connection.sendResponse(this)}}
to {{connection.sendResponse(call)}}.
> Design Server.Call to be extensible for unified call queue
> ----------------------------------------------------------
>
> Key: HADOOP-13465
> URL: https://issues.apache.org/jira/browse/HADOOP-13465
> Project: Hadoop Common
> Issue Type: Improvement
> Components: ipc
> Reporter: Daryn Sharp
> Assignee: Daryn Sharp
> Attachments: HADOOP-13465.1.patch, HADOOP-13465.patch
>
>
> The RPC layer supports QoS but other protocols, ex. webhdfs, are completely
> unconstrained. Generalizing {{Server.Call}} to be extensible with simple
> changes to the handlers will enable unifying the call queue for multiple
> protocols.
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
