Hadoop - stop-dfs.sh, stop-mapred.sh does not stop all instances

[Grandl Robert <rg...@yahoo.com>]

Hi all,

Probably is a stupid question but I don't figure out what happens.

For example I start on master, start-dfs.sh, start-mapred.sh

However, when I want to stop: stop-mapred, stop-dfs, only one slave instance is killed. 

For example in on the master I am doing something like that:

rgrandl@ikq01:~/hadoop/hadoop-0.21.0$ bin/stop-mapred.sh 
stopping jobtracker
ikq02.rrr: no tasktracker to stop
ikq03.rrr: stopping tasktracker

But I have a tasktracker on ikq02 as well which is not stopped anyway.
rgrandl@ikq02:~$ jps
2462 Jps
2211 TaskTracker
2109 DataNode

Same thing for stop-dfs.sh


Always the tasktracker, datanode instances are killed only on one slave node, even if I have maybe 20 slaves or more. 

Thanks for any clue,
Robert





      

Re: Hadoop - stop-dfs.sh, stop-mapred.sh does not stop all instances

[Konstantin Boudnik <co...@boudnik.org>]

Without much details I only can offer to check if your master hosts (NN, JT)
have passwordless access to all slaves

Cos

On Thu, Nov 11, 2010 at 02:41AM, Grandl Robert wrote:
> Hi all,
> 
> Probably is a stupid question but I don't figure out what happens.
> 
> For example I start on master, start-dfs.sh, start-mapred.sh
> 
> However, when I want to stop: stop-mapred, stop-dfs, only one slave instance is killed. 
> 
> For example in on the master I am doing something like that:
> 
> rgrandl@ikq01:~/hadoop/hadoop-0.21.0$ bin/stop-mapred.sh 
> stopping jobtracker
> ikq02.rrr: no tasktracker to stop
> ikq03.rrr: stopping tasktracker
> 
> But I have a tasktracker on ikq02 as well which is not stopped anyway.
> rgrandl@ikq02:~$ jps
> 2462 Jps
> 2211 TaskTracker
> 2109 DataNode
> 
> Same thing for stop-dfs.sh
> 
> 
> Always the tasktracker, datanode instances are killed only on one slave node, even if I have maybe 20 slaves or more. 
> 
> Thanks for any clue,
> Robert
> 
> 
> 
> 
> 
>