LHC Computer Grid - Rollout
> [mailto:[log in to unmask]] On Behalf Of David Smith said:
> Yes, it is published but no it's not used. Infact I can open a bug for
> this - although there may be reasons why doing the test at
> matchmaking is problematic.
Well, you might have race conditions, but I think the main reason is that
the RB itself doesn't know anything at all about information in the schema,
it only acts on what it finds in the Requirements and Rank in the JDL -
which could include a job limit just as it includes the "queue active" flag,
but only if someone decides to do it. Doing things that way makes it
flexible, but I think it's had the unfortunate side effect that it puts the
onus on the users to use the right expressions (or the sysadmins for the
default values), and the result is that things are often not used in the
optimum way.
Stephen
|