hadoop-common-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From "lixiangna (JIRA)" <j...@apache.org>
Subject [jira] Commented: (HADOOP-3499) decommission node is both in the "Live Datanodes" with "In Service" status, and in the "Dead Datanodes" of the dfs namenode web ui.
Date Wed, 11 Jun 2008 10:58:46 GMT

    [ https://issues.apache.org/jira/browse/HADOOP-3499?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=12604179#action_12604179
] 

lixiangna commented on HADOOP-3499:
-----------------------------------

Hi,

I found where the problem is:
I gave the hostname of decommission nodes in the excludesFile, but
those hostnames were not the full name(no xx.com suffix). the namenode
would look about if a datanode's name(ip:port) or host(ip) or hostname
contained by the excludelist. And the hostname is the full name, not
equal to the name in the excludelist, so nodes weren't considered in
the excludelist. Decommission had never successed actually!  the nodes
in excludesFile were seemed as other new nodes.

I guess if converting hostname of exclude node to ip for compare will
be better. Because one ip can have lots hostname. anyhow, no change is
also ok. if so the bug could be end?

thanks



> decommission node is both in the "Live Datanodes" with "In Service" status, and in the
"Dead Datanodes" of the dfs namenode web ui.
> -----------------------------------------------------------------------------------------------------------------------------------
>
>                 Key: HADOOP-3499
>                 URL: https://issues.apache.org/jira/browse/HADOOP-3499
>             Project: Hadoop Core
>          Issue Type: Bug
>          Components: dfs
>    Affects Versions: 0.17.0
>         Environment: linux-2.6.9
>            Reporter: lixiangna
>   Original Estimate: 72h
>  Remaining Estimate: 72h
>
> try to decommission a node by the following the steps:
> (1) write the hostname of node which will be decommissioned in a file (the exclude file)
> (2) specified the absolute path of the exclude file as a configuration parameter dfs.hosts.exclude.
> (3) run "bin/hadoop dfsadmin -refreshNodes".
> It is surprising that the node is found both in the "Live Datanodes" with "In Service"
status, and in the "Dead Datanodes" of 
> the dfs namenode web ui. When copy new data to the HDFS, its Used size is increasing
as other un-decommissioned nodes. 
> Obviously it is in service. Restarting the HDFS or waiting a long time(two day) havn't
make the decommission yet.
> the more strange thing, If nodes are configured as the include nodes by similar steps,
then these include nodes and
> the exclude node are all only in the "Dead Datanodes" lists. 
> I did many times tests in both 0.17.0 and 0.15.1. The results is same. So i think there
maybe bugs.

-- 
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