hadoop-common-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From "Raghu Angadi (JIRA)" <j...@apache.org>
Subject [jira] Commented: (HADOOP-1261) Restart of the same data-node should not generate edits log records.
Date Mon, 16 Apr 2007 19:49:15 GMT

    [ https://issues.apache.org/jira/browse/HADOOP-1261?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel#action_12489215
] 

Raghu Angadi commented on HADOOP-1261:
--------------------------------------

> Could you check whether we can also get rid of clusterMap.remove() and clusterMap.add()
in case nodeN == nodeS. 

Then we also need to make sure that the node is on the same rack etc. I am not sure if the
extra code is worth it. Node registration is not supposed to be very often. Simpler it is,
the better.

> Also I see dnReg is used only to set a correct ip address in nodeReg. There should be
an easier way.

hm. will check. Currently it just uses existing interface.


> Restart of the same data-node should not generate edits log records.
> --------------------------------------------------------------------
>
>                 Key: HADOOP-1261
>                 URL: https://issues.apache.org/jira/browse/HADOOP-1261
>             Project: Hadoop
>          Issue Type: Bug
>          Components: dfs
>    Affects Versions: 0.12.0
>            Reporter: Konstantin Shvachko
>         Assigned To: Raghu Angadi
>            Priority: Minor
>             Fix For: 0.13.0
>
>         Attachments: HADOOP-1261-1.patch
>
>
> Currently during registration of a data-node with the storage id that has already been
registered
> the name-node logs 2 records: <add node1> <remove node2>.
> If the same node has been restarted these two records are redundant, lead to unnecessary
> increase of the edits file and result in a slow-down of the name-node startup.

-- 
This message is automatically generated by JIRA.
-
You can reply to this email to add a comment to the issue online.


Mime
View raw message