Burke, S (Stephen) wrote:
> LHC Computer Grid - Rollout
> > [mailto:[log in to unmask]] On Behalf Of James Casey said:
> > No, the behaviour when we use a certificate is the same. i.e. the VOMS
> > functionality is in there. It hasn't been tested in a while, but it
> > should work.
>
> I don't see how that's possible. With the old behaviour, if I had a dteam
> VOMS proxy and I did "edg-rm pi" it would be the same as doing "edg-rm
> --vo=dteam pi" without the VOMS proxy. With your change "edg-rm pi" should
> mean "all Vos". I don't see how it can do both!
>
Ah yes, I see what you mean. If you have a VOMS proxy, you'll only get
your VO. With a non-VOMS proxy, you'll get all VOs.
> > > Incidentally, this ought to be a silly question, but is
> > it guaranteed
> > > that
> > > the bdii used by the broker for matchmaking is the same one
> > used by the RM
> > > when called by the broker to look up input files?
> > >
> > No, it's not! Currently the RM uses the BDII specified in the host
> > specific config file
> > /opt/edg/var/etc/edg-replica-manager/edg-replica-manager.conf.
>
> You misunderstood my point, I was asking whether the bdii used for the RM
> *on the RB* has to be the same as the one used for matchmaking on the same
> RB. Are they configured independently? Obviously you would expect them
> to be
> the same anyway, but if they are separate settings it leaves some scope for
> things to go wrong.
No, they should both use the edg-replica-manager.conf file. The BDII
uses the C++ replica-manager client libraries, which read the
edg-replica-manager.conf file.
James.
|