[
https://issues.apache.org/jira/browse/HADOOP-993?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel#action_12481728
]
Yoram Arnon commented on HADOOP-993:
------------------------------------
I'm not sure how an include/exclude file helps: it will contain the
administrator's intent, but won't show which nodes are actually there and which
died or haven't yet started - a key factor after a restart or upgrade.
I'm all for simplicity, but not at the expense of functionality.
If the patch could include, or follow, an alternative method for the existing
functionality I'd be happier.
Also, by not tracking datanodes in the namenode, aren't we precluding future
enhancements, such as determining the 'true' owner of a storage id? Even
currently, when the namenode assigns a new datanode a storage id, it compares
against all existing storage ids. If the information isn't persistent, it might
allocate an existing id to a new node, just because it connected before the
owner of the id connected.
> Namenode does not need to store any data node info persistently.
> ----------------------------------------------------------------
>
> Key: HADOOP-993
> URL: https://issues.apache.org/jira/browse/HADOOP-993
> Project: Hadoop
> Issue Type: Improvement
> Components: dfs
> Reporter: Raghu Angadi
> Assigned To: Sameer Paranjpye
> Attachments: noDatanodesInFsimage.patch
>
>
> Namenode does not need to serialize datanode info. It will map datanode to
> storageID when datanode register.
--
This message is automatically generated by JIRA.
-
You can reply to this email to add a comment to the issue online.
