On Tue, Jun 2, 2009 at 11:38 PM, Maarten Litmaath
<[log in to unmask]> wrote:
> Hi Kashif,
>
>> I am trying to install a CE which can submit job to separate pbs cluster
>> and created a small test setup.
>> I installed a separate torque cluster with one headnode, a separate CE
>> and tar-WN on a shared area. All these machines share home directories.I
>> am using pbs as jobmanager in CE.
>> Pool account was created through yaim at CE and then copied content of
>> /etc/groups,passwd and shadow file on all the machines, so pool account
>> exist on all the machine. Since all machine are using the same home
>> directory so I thought there is no need of passwdless ssh between wn and
>> ce.
>>
>> My question is that how ce transfer userproxy to wn,s. I found here
>> http://goc.grid.sinica.edu.tw/gocwiki/Unspecified_gridmanager_error that
>> lcg jobmanager uses #PBS stagein<proxy> to transfer userproxy to wn. Is
>> it correct?
>
> That is roughly correct for the "lcgpbs" (!) job manager.
>
>> Although home directory is shared but do I have to still setup password
>> less ssh between ce and wn.
>
> The "pbs" job manager does not need that.
It does not need it but depends how you set things up, In order not
to have ssh needed you must have
$usecp *:* *
(Check my synxtax) In your /var/spool/pbs/mom_priv/config
scp is still the default reguardless of if mounts are actually in place.
Steve
>
--
Steve Traylen
|