[GE users] ssh_exchange_identification

Reuti reuti at staff.uni-marburg.de
Wed Aug 8 21:47:31 BST 2007


Am 08.08.2007 um 16:31 schrieb Mathias Goldau:

> * I've had trouble with my qrsh installation so I redirected qrsh  
> to use
> ssh instead of telnet. Now host-keys are working and all qrsh, qlogin
> stuff works as root well, but as ordinary user don't.
>
> * As ordinary user I can't do a "qrsh hostname", but as root all is
> well. The Error message I provide again is short:
>
> -sh-3.00$ qrsh -l hostname=node23 -verbose hostname
> local configuration frontend not defined - using global configuration
> Your job 160386 ("hostname") has been submitted
> waiting for interactive job to be scheduled ...
> Your interactive job 160386 has been successfully scheduled.
> Establishing /usr/bin/ssh session to host node23 ...
> qrsh_starter: executing child process (null) failed: No such file or
> directory
> /usr/bin/ssh exited with exit code 0
> reading exit code from shepherd ... 1

You mean:

qrsh /bin/hostname

is giving the same "No such file or directoty" error message from  
qrsh_starter? At least: your ssh setup seems working, as otherwise  
the qrsh_starter wouldn't come up.

OTOH: If you ssh by hand to this node, the "hostname" command is  
working for these normal/system users?

> * I know almost two workarounds which work as ordinary user (but this
> doesn't statisfies me nor my boss :/)
>
>  1. QRSH_WRAPPER=env qrsh hostname

Hence the command env is found on the node. Is giving:

qrsh /bin/echo \$PATH

saying something?

>  2. qrsh -noshell tcsh -f -c "hostname"
>
> * I tried this with absolute path too, but no chance, also I made an
> strace of qrsh some posts before, but this looks really cryptic too  
> me.
>
>
>> - By default, qlogin is also defined if there is a local  
>> configuration
>> for the node. There is none in your cluster? 	
>
> Yes. My cluster configuration (seen with qmon) has just a global  
> cluster
> configuration but there qlogin is defined:
>
> $ qconf -sconf | grep qlogin
> qlogin_command               /usr/bin/ssh_qlogin_wrapper
> qlogin_daemon                /usr/sbin/sshd -i
>
> $ cat /usr/bin/ssh_qlogin_wrapper
> #!/bin/sh
> HOST=$1
> PORT=$2
> /usr/bin/ssh -X -p $PORT $HOST
>
>
>> - Can you please check in the messages file of the node, whether sshd
>> was really set up in the last change of the configuration?
>
> I did a "tail -f /var/log/messages" if you mean that and got the  
> following:
>
> Aug  8 16:23:26 node23 sge_shepherd-160392: nss_ldap: reconnecting to
> LDAP server...
> Aug  8 16:23:26 node23 sge_shepherd-160392: nss_ldap: reconnected to
> LDAP server after 1 attempt(s)
> Aug  8 16:23:27 node23 kernel: SELinux: initialized (dev 0:1a, type
> nfs), uses genfs_contexts
> Aug  8 16:23:27 node23 sshd(pam_unix)[11149]: session opened for  
> user me
> by (uid=0)
> Aug  8 16:23:27 node23 sshd(pam_unix)[11149]: session closed for  
> user me

No, I meant $SGE_ROOT/default/spool/node23/messages

After you change the SGE configuration with qconf -mconf, it will  
update to reflect the actual settings.

> could it be that my sge_shepherd is configured to work only with ldap?
> My ordinary user isn't an ldap user. It is just a system user in

What is in /etc/nsswitch.conf? It is possible to rely on nis and then  
skip the local files at all.

> /etc/passwd and so on.
>
>
>> - While waiting for the return of "qrsh hostname": can you login  
>> to the
>> node and check with "ps -e f" whether there was anything started  
>> by the
>> shepherd?
>
> headnode: qrsh -l hostname=node23 -verbose hostname
> node23:   watch -n 0 "ps -e f | grep shepherd"

Yes, thx - but this way we don't see any kids of the shepherd... Just  
copy the relevant lines from a "ps f -eo  
pid,ppid,pgrp,user,ruser,command"

-- Reuti


> output from node23's watch command:
> 6578 pts/0 S+  0:00     \_ watch -n 0 ps -e f | grep shepherd
> 7790 pts/0 S+  0:00         \_ sh -c ps -e f | grep shepherd
> 7792 pts/0 S+  0:00             \_ grep shepherd
> 7785 ?     S   0:00  \_ sge_shepherd-160391 -bg
> 7786 ?     Ss  0:00      \_ sge_shepherd-160391 -bg
>
> thanks a lot for your great help.
>
> ---------------------------------------------------------------------
> To unsubscribe, e-mail: users-unsubscribe at gridengine.sunsource.net
> For additional commands, e-mail: users-help at gridengine.sunsource.net

---------------------------------------------------------------------
To unsubscribe, e-mail: users-unsubscribe at gridengine.sunsource.net
For additional commands, e-mail: users-help at gridengine.sunsource.net




More information about the gridengine-users mailing list