Testbed Support for GridPP member institutes [mailto:TB-
> [log in to unmask]] On Behalf Of Stephen Jones said:
> On 06/18/2015 03:43 PM, Stephen Burke wrote:
> > The vast majority of problems come from the CE batch system adapters.
>
> I assume you mean GIP Plugins and so on.
Specifically the things which query the batch system for attributes like the number of running and queued jobs per VO. That's usually very fragile because it tends to reply on things like the precise output format from a command which aren't explicitly documented and can change unpredictably between versions. It can also be sensitive to exactly how the batch system is configured, which also varies widely. The SE publishers are also complex, but they're tightly bound to the implementations so it's easier for the developers to maintain them.
> Yes, it's true. I think a large part of the problems we've had
> came from there. IMO those implementations could do with
> a good going over.
They did in fact get quite a big revision in EMI - I don't have a good sense of whether the number of problems has reduced as a result.
> I personally struggle with the data semantics of the content. Maybe
> I need someone to mentor me in the meaning of the tags: where did
> they come from, what's going to happen to them and what are they
> all about?!?
I'm not really sure what you're asking - the schema has a history going back at least to the start of Datagrid in 2001 so that's potentially an awful lot of information to explain!
Stephen
|