On 22/08/2012 15:24, Stephen Jones wrote:
> On 08/22/2012 01:58 PM, John Hill wrote:
>> I've just added (it appears successfully) a couple of extra WNs to the
>> Cream CE on our site. I updated the values of
>>
>> CE_PHYSCPU
>> CE_LOGCPU
>> CE_SI00
>> CE_CAPABILITY
>> CE_OTHERDESCR
>>
>> in site-info.def on the CE and re-yaimed. Is there anywhere else I
>> need to insert the updated information, or will it eventually
>> propagate everywhere?
>>
>> John
>
> Hi John,
>
> This list looks like it might have come from
> https://www.gridpp.ac.uk/wiki/Publishing_tutorial, e.g.
>
> CE_PHYSCPU=174
> CE_LOGCPU=952
> CE_CAPABILITY="CPUScalingReferenceSI00=2500 Share=atlas:63
> Share=lhcb:25 glexec"
> CE_SI00=3653
> CE_OTHERDESCR=Cores=5.47,Benchmark=14.61-HEP-SPEC06
>
>
> Those are all the things you need to change on an existing CE to set the
> power and accounting publishing.
>
> Yaim pushes them into the appropriate place for the BDII to pick them up
> and transmit them onwards.
>
> If the nodes run jobs (i.e. torque knows about them) then I think that's
> it.
>
> IIRC I don't do any more than that when we add new nodes.
>
> Steve
>
> PS: do you use a scaling factor on your worker nodes, or all all your
> nodes equal?
>
Hi Stephen,
I certainly used that URL as a guide as to what needed changing -
thanks. Santanu never set up scaling factors for the different CPUs - I
will look at that sometime - but not in the immediate future.
It appears that my reconfiguration of the Cream CE has broken job
submission - though I can't see why. There seem to be no obvious errors
in the log files (which presumably means that I haven't found the right
one yet :-( ).
John
