[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]
Re: 2-tape drive testing
Luis Bernardo wrote:
>
> On Thu, 1 Oct 1998, David M. Malon wrote:
>
> > >
> > > I was thinking a little more about the 2-tape drive issue.
> > > Rather than arange to have 2 drives put into one class
> > > of service so that an individual tape may go into either
> > > drive, what about just spreading files over 2 existing
> > > classes of service. When we retrieve them, a given file
> > > will always go to the same tape drive, but we will
> > > know that the files are spread across at least 2 tapes.
> > > We can also do this without a reconfiguration of hpss.
> > >
> > > Doug
> > >
> >
> > Wouldn't this imply that Luis would need to know the class of
> > service, rather than just the HPSS path, for every file? I know he uses
> > setcos with a fixed number in his pftp scripts now. If Luis
> > needs to write code to accommodate this approach, it becomes a matter of
> > resources and time (GCA effort vs RCF effort?); I think Luis is also
> > working on a couple of other issues (we've been running with cmanager.old,
> > for example).
>
> I believe the problem with cmanager (not cmanager.old) was the script. I
> changed "cmanager.old -g $1" into "cmanager $1" and it's running now.
> Alex, what's that -g for? cmanager was not expecting that and that's why
> it was crashing. That's what I think.
>
> >
> > Luis, what would implementing Doug's suggestion require on your part?
> >
> In principle it would be quick to make changes to accommodate different
> classes of service and different HPSS paths for different files. The issue
> is where the CM would get that information from.
>
> Luis
>
> --------------------------------------------------------------
> Luis M. Bernardo <lmbernardo@lbl.gov> <www.lbl.gov/~bernardo>
> Scientific Data Management Research & Development Group
> Lawrence Berkeley National Laboratory
> Berkeley, CA 94720
>
When the cmanager crashed, the script was not used.
In fact, the script uses only fake cache manager at this point.
I never made the script use the real cmanager.
So, crashing comes from somewhere else. In fact, when the crash
happened, cmanager read all the gc.config information and then gave a
segmentation fault.
The reason that there is -g in the cmanager.sh (script) is that fake
cache manager uses it
for the path for gc.config.
--Alex
--
Alex Sim
MS 50B-3238 http://sim.lbl.gov
DM, NERSC, LBNL ASim@lbl.gov
1 Cyclotron Road office: (510) 495-2290
Berkeley, CA 94720 FAX: (510) 486-4004