Hi,
Thanks Peter.
Mixing EMI-2 and EMI-3 WNs is an interesting use case for us, because it
avoids a complete downtime of the site for the upgrade, though we will
certainly schedule one if it is required.
However, since we do not support MPI jobs at CIEMAT, let me ask for more
information regarding your comment: are there any other specific
problems related to the mix of WN from different versions, or just MPI?
Well, even if you tell me that there are no precise issues but that the
configuration was not tested or is in general suspected to cause
problems, we will consider it enough to avoid risking it :-)
Cheers,
Antonio
> Dear Antonio,
>
> EGI UMD verifiers tested, although not extensively, a mixed
> configuration with CREAM and WN from different EMI versions.
> CREAM EMI-3 and WN EMI-2 are compatible. And also the other way around.
>
> What is not advisable is to have some WN from EMI-2 and other from
> EMI-3 in the same subcluster, because some applications such as MPI do
> not work properly with different WN versions together.
>
> Regards
> Peter
>
>
> On 11 February 2014 18:47, Antonio Delgado Peris
> <[log in to unmask]> wrote:
>> Dear all,
>>
>> We have recently received a message regarding the UMD-2/EMI-2
>> decommissioning calendar, notifying us about the need to start planning the
>> upgrade to EMI/UMD-3 services.
>>
>> Thinking about a possible gradual upgrade, I would like to ask about
>> compatibility issues when mixing EMI/UMD-2 and EMI/UMD-3 services, in
>> particular in what regards WNs, glexec, CREAMs, Argus and BDII. E.g.: can we
>> mix different WNs from both releases and different CREAMs from both also?
>>
>> Given that I've found no warning against this, my first idea is that all
>> combinations should be possible (except for the case of APEL**), but I would
>> like to be sure.
>>
>> ** Warned here: https://wiki.egi.eu/wiki/UMD-3:UMD-3.0.0#emi.apel.sl6.x86_64
>>
>> Thanks a lot for any comments!
>>
>> Cheers,
>>
>> Antonio
>
>
|