LHC Computer Grid - Rollout
> [mailto:[log in to unmask]] On Behalf Of Henry Nebrensky
said:
> My point was that if an SGM is assumed to be optional, then
> isn't Juan's
> users.conf technically valid - albeit with one eccentrically
> named pool
> user? How then can YAIM claim an error?
As I already pointed out, the site-info.def defines an LDAP endpoint for
an sgm group, which seems to imply that you expect it to exist. (Maybe
someone should see what happens if it's omitted?)
> Requiring every VO to have an SGM (with exactly one pool
> account?)
The sgm accounts are *not* pool accounts, they are statically mapped.
That's what has caused the problem here ...
> It's probably safer than assuming that e.g. all pool accounts are
> (decimally) numbered...
Pool accounts *must* be decimally numbered, that's how the system works.
There are two formats, a mapping of a DN to e.g. "dteamsgm" is a fixed
mapping to an account called "dteamsgm", whereas a mapping to ".dteam"
maps to accounts called "dteamnnn" where nnn is a number. The reason for
the failure in this case is that the generated mapping said ".dteamsgm"
and there are no accounts called "dteamsgm001" etc.
Stephen
|