Olivier van der Aa wrote:
> Graeme Stewart wrote:
>
>
> Hi Graeme,
>
>> This is very wrong: "se01.htc.esc.qmul".
>>
>
> The machine as two ethernet interface and this is the name of the
> interface on the private net.
> The setup is special since the pool is on the admin node with the 20Tb
> mounted on it.
OK, this might make things interesting!
When I did dual homing in Glasgow it was very, very important to ensure
that the admin node and pool nodes have the _same_ hostname when
resolved inside and outside [1]. So, se2-gla.scotgrid.ac.uk has to be
10.0.5.5 (from a worker node), but 194.36.1.113 from any extrenal host.
>
> It seems that gridftp start to listen to the private interface rather
> than the public one.
Try netstat -tlp - you shoudl find that gridftp is bound to all
interfaces - again see [1].
>
>> There is a limitation in DPM at the moment that pools must be in the
>> same domain as the admin node (it's a rfio problem - will be fixed).
>>
>> Did you add the filesystem with the simple hostname?
>>
>
> sorry what do you mean.
When you add a filesystem on, say, node1.domain.ac.uk:/store, then you
need to do:
# dpm-addfs --pool poolName --server node1 --fs /store
_not_
# dpm-addfs --pool poolName --server node1.domain.ac.uk --fs /store
What's the output of dpm-qryconf?
graeme
[1] http://wiki.gridpp.ac.uk/wiki/DPM_Dual_Homing
--
--------------------------------------------------------------------
Dr Graeme Stewart http://www.physics.gla.ac.uk/~graeme/
GridPP DM Wiki http://wiki.gridpp.ac.uk/wiki/Data_Management
|