[
https://issues.apache.org/jira/browse/HADOOP-7380?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]
Aaron T. Myers updated HADOOP-7380:
-----------------------------------
Attachment: hadoop-7380.0.patch
Here's a first hack at the Common portion of client failover, not intended for
commit.
If people think this approach is roughly sound I'll fix it up and prepare it
for commit. It still needs at least the following work:
* A way to mark certain interface methods as idempotent (and thus should be
retried upon failover). This will probably take the form of a method annotation.
* Refinement of the precise set of exceptions to deal with.
* Potential refinement of the {{FailoverProxyProvider}} interface. I'm not in
love with it as it stands.
I've also created a test which uses this code to actually perform a failover
between two federated NNs (which happen to have the same FS metadata) by
starting them both up, creating a file in each at the same path, shutting down
one, and ensuring the DFSClient properly fails over to the second. I can post
that patch on HDFS-1973 but it depends on this patch.
> Common portion of HDFS-1973
> ---------------------------
>
> Key: HADOOP-7380
> URL: https://issues.apache.org/jira/browse/HADOOP-7380
> Project: Hadoop Common
> Issue Type: New Feature
> Components: ipc
> Affects Versions: 0.23.0
> Reporter: Aaron T. Myers
> Assignee: Aaron T. Myers
> Fix For: 0.23.0
>
> Attachments: hadoop-7380.0.patch
>
>
> Implementing client failover will likely require changes to {{o.a.h.io.ipc}}
> and/or {{o.a.h.io.retry}}. This JIRA is to track those changes.
--
This message is automatically generated by JIRA.
For more information on JIRA, see: http://www.atlassian.com/software/jira
