[Wien] problems with ssh/csh (or so) on a cluster

[Peter Blaha]

> Under "host nodes for w2web" it says: "URL including http(s) and port to host-nodes", so I specified
> http://node.chemie.uni-mainz.de,

Sure, for w2web you need an URL, but not for ssh.

> ssh node is working without password, as is needed.

Does it open a new window on your screen or does it stay in the original terminal window ?
Your error message indicates that the ssh process wants to open a new window and of
course this does not work in a "batch-job".
It MUST NOT open a new window, if it does, you have to reconfigure your ssh defaults.

can you do:

ssh node.chemie.uni-mainz.de hostname

or
ssh node.chemie.uni-mainz.de "lapw1 lapw1.def"
(This is almost what is executed in the parallel scripts.

Last tip: check your startup files (.cshrc, .bashrc ,...) to be sure that
there is nothing forcing you to have a X-window environment (like a xbiff statement or
similar).

> 
> I changed my .machines file to
> 
> 1:node.chemie.uni-mainz.de
> 1:node.chemie.uni-mainz.de
> granularity:1
> extrafine:1
> lapw0:node.chemie.uni-mainz.de
> 
> when omitting the lapw0:node.chemie.uni-mainz.de in the machines file, lapw0 is executed and the crash occurs at lapw1, but with another error message in STDOUT:
> Error: Can't open display: localhost:22.0
> Host key verification failed.
> cat: No match.
> And in the console "X11 connection rejected because of wrong authentication." appears. As I said, the key is appended and ssh node is working withour password.
> 
> 
> 
> ________________________________________
> Von: wien-bounces at zeus.theochem.tuwien.ac.at [wien-bounces at zeus.theochem.tuwien.ac.at] im Auftrag von Peter Blaha [pblaha at theochem.tuwien.ac.at]
> Gesendet: Mittwoch, 17. September 2008 08:05
> An: A Mailing list for WIEN2k users
> Betreff: Re: [Wien] problems with ssh/csh (or so) on a cluster
> 
> You should not specify an URL, but just a qualified hostname.
> 
> You must be able to do on a terminal window:
> 
> ssh hostname     or ssh hostname.domain  (like
> "homer.physik.uni-mainz.de" or whatever the full hostname is)
> 
> and get connected without a password.
> 
> Please read the UG about parallelization. Your .machines file does not
> make sense. You need at least 2 lines
> 
> 1:host1
> 1:host2
> 
> to get a 2-node k-point parallelization (on dual or quadcores, host2 can
> be equal host1)
> 
> The parallelization of
> lapw0: host1 host2      needs mpi !!!
> 
>> There is no problem with a non-parallel calculation, but as soon as I switch to parallel, I get the following error message:
>>
>> Pseudo-terminal will not be allocated because stdin is not a terminal.
>> ssh: cd /home/ludwigc/Wien2k/test_128;time lapw0 lapw0.def: Name or service not known
>>
>>>   stop error
>> I use execution type custom = %f, the URL of the host nodes is given. Machines file is
>>
>> 1:<URL host node>
>> granularity:1
>> extrafine:1
>> lapw0:<URL host node>
>>
>> The rsa key is appended to authorized_keys. If you need any more information to help, please tell me.
>>
>> All the best,
>> Christian Ludwig
>>
>> _______________________________________________
>> Wien mailing list
>> Wien at zeus.theochem.tuwien.ac.at
>> http://zeus.theochem.tuwien.ac.at/mailman/listinfo/wien
> _______________________________________________
> Wien mailing list
> Wien at zeus.theochem.tuwien.ac.at
> http://zeus.theochem.tuwien.ac.at/mailman/listinfo/wien
> _______________________________________________
> Wien mailing list
> Wien at zeus.theochem.tuwien.ac.at
> http://zeus.theochem.tuwien.ac.at/mailman/listinfo/wien

-- 

                                       P.Blaha
--------------------------------------------------------------------------
Peter BLAHA, Inst.f. Materials Chemistry, TU Vienna, A-1060 Vienna
Phone: +43-1-58801-15671             FAX: +43-1-58801-15698
Email: blaha at theochem.tuwien.ac.at    WWW: http://info.tuwien.ac.at/theochem/
--------------------------------------------------------------------------