> -----Original Message-----
> From: Testbed Support for GridPP member institutes [mailto:TB-
> [log in to unmask]] On Behalf Of Matt Doidge
> Sent: 18 January 2016 15:40
>
> NGI
> https://ggus.eu/?mode=ticket_info&ticket_id=118930 (18/1)
> The NGI received a ticket concerning incorrect or missing glue
> information for the Tier 1, Brunel, Imperial, Liverpool, Durham,
> Glasgow, Bristol, Oxford and RALPP.
> The variables in question are GlueSubClusterPhysicalCPUs,
> GlueSubClusterLogicalCPUs and GlueHostProcessorOtherDescription. There
> are some extra instructions in the ticket - it would be nice if we
> didn't have to create child tickets (hint hint...).
>
Let's just close it; it's complete bullshit. There is no way we're ever going to publish information that accurately describes the actual hardware in the way glue requires, with a full subcluster definition for every single batch of worker nodes in a cluster, and there's no reason we should. These tickets amount to "please lie in a manner the automatic tests can't easily spot", and that's just not an interesting use of anyone's time.
I'm not sure if we can publish strings in these fields, but if we can I'm all for adopting "GlueCEMindYourOwnDamnBusiness" as a standard value.
Ewan
|