On Fri, 22 Nov 2002, Dr D J Colling wrote:
> I must confess that the whole area of replica catalogues and management is
> the area of edg that I know least about. Also I am teaching all morning
> but will try to investigate this afternoon. The RB is certainly on a 1.2
> machine. I have forwarded this mail to the tb-support in the hope that
> some there who knows more about these things can give me some advice as to
> where to start looking.
I'll try ...
> > InputData = {"LF:job-status03"};
> > ReplicaCatalog = "ldap://bfb.hep.man.ac.uk:389/lc=file
> > collection,rc=BaBarReplicaCatalog,dc=gridpp,dc=ac,dc=uk";
> > DataAccessProtocol = "file";
If I look at this URL:
ldap://bfb.hep.man.ac.uk:389/lc=file collection,rc=BaBarReplicaCatalog,dc=gridpp,dc=ac,dc=uk
it looks OK to me - there is a space in the collection name which is
OK but you have to make sure it doesn't get lost (it may get wrapped
in my mail and has been in yours I guess).
The location objects are called GlobusReplicaInfo which tells me
that it's a 1.2-style RC, the new 1.3 release will need the RC
to be recreated and the location objects will then be called
GlobusReplicaLocation. You are presumably using a fully 1.2-system -
1.2 and 1.3 are totally incompatible in this respect so that's an
important thing to check. If you use a 1.3 system to write to a 1.2
catalogue it can get corrupted, but I don't see any sign of that here.
The LFN job-status03 is indeed registered (apparently with the
replica manager rather than GDMP) and as you say is recorded at Lyon,
Slac and RAL.
One thing I might wonder about is whether the RC machine was unreachable
at the time. I can certainly see it now; is the problem repeatable?
One other thing to try is taking the lc= part out of the RC URL in the
JDL; some versions of the JDL require it but I think older ones did
not. So try:
ReplicaCatalog = "ldap://bfb.hep.man.ac.uk:389/rc=BaBarReplicaCatalog,dc=gridpp,dc=ac,dc=uk";
In general it probably isn't worth chasing this too hard given that
everything has to change for 1.3 anyway ... I can't test it myself because
the main 1.2 production system was shut down yesterday.
Stephen
|