[jira] [Updated] (HADOOP-11518) Prefix-based per-RPC server configuration

["Kihwal Lee (JIRA)" <ji...@apache.org>]

     [ https://issues.apache.org/jira/browse/HADOOP-11518?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel ]

Kihwal Lee updated HADOOP-11518:
--------------------------------
    Attachment: HADOOP-11518.patch

The patch adds {{PrefixedConfiguration}} class which overrides the base {{get()}} method for prefixed value retrieval and the fall-back behavior.  New static IPC call queue config refresh methods have been added to {{RPC}}. Prior to this, {{NameNodeRpcServer}} was directly talking to the server instances for refresh.

> Prefix-based per-RPC server configuration
> -----------------------------------------
>
>                 Key: HADOOP-11518
>                 URL: https://issues.apache.org/jira/browse/HADOOP-11518
>             Project: Hadoop Common
>          Issue Type: Improvement
>          Components: ipc
>            Reporter: Kihwal Lee
>            Assignee: Kihwal Lee
>         Attachments: HADOOP-11518.patch
>
>
> If a process runs multiple RPC servers, there is no way to individually set their IPC parameters beyond what's supported by the RPC Builder.  If a configuration is shared among many nodes, this is more problematic. 
> In this jira, a prefix-based per-server configuration is proposed.  When an RPC server is built, a prefix can be specified, which will then be prepended to keys when retrieving configured values.  There will be an option to fall-back to using the key without the prefix when no value is found with the prefix, thus making this feature compatible with the current way of configuring servers.



--
This message was sent by Atlassian JIRA
(v6.3.4#6332)