[
https://issues.apache.org/jira/browse/HDFS-13857?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16594062#comment-16594062
]
Íñigo Goiri commented on HDFS-13857:
------------------------------------
Thanks [~hfyang20071] for the patch.
I'm not sure the implementation in [^HDFS-13857.001.patch] is onyl for writes
as it will also disallow reads.
I'm fine with that but we need to rename everything accordingly.
Some other comments:
* We could throw the exception in the lookup itself and then just add if
(!defaultNSWriteEnable) throw new IOException("Cannot find locations for " +
path);
* The message could be a little more specific in this case and say we don't
support this because the default is not enabled.
* We should have the false unit test too.
* The change with FEDERATION_MOUNT_TABLE_MAX_CACHE_SIZE should be avoided to
reduce churn.
> RBF: Choose to enable the default nameservice to write files.
> -------------------------------------------------------------
>
> Key: HDFS-13857
> URL: https://issues.apache.org/jira/browse/HDFS-13857
> Project: Hadoop HDFS
> Issue Type: Improvement
> Components: federation, hdfs
> Affects Versions: 3.0.0, 3.1.0, 2.9.1
> Reporter: yanghuafeng
> Assignee: yanghuafeng
> Priority: Major
> Attachments: HDFS-13857.001.patch
>
>
> The default nameservice can provide some default properties for the namenode
> protocol. And if we cannot find the path, we will get a location in default
> nameservice. From my side as cluster administrator, we need all files to be
> written in the location from the MountTableEntry. If no responding location,
> some error will return. It is not better to happen some files are written in
> some unknown location. We should provide a specific parameter to enable the
> default nameservice to store files.
--
This message was sent by Atlassian JIRA
(v7.6.3#76005)
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
