Matt,
Are there any relevant error messages in /var/log/pnfsd.log ? Like
Derek, I've also seen something like this when DB's don't get started
properly. What number of shmservers are defined in the
/usr/etc/pnfsSetup file? Is this larger than the number of DBs that you
require?
Cheers,
Greig
Matt Doidge wrote:
> Hello,
> I'm almost done bringing our dcache fully back to life. We've got
> working transfers and we're passing the rm tests, we've just got to
> get the gridpp GIP sorted. I spent the day puzzling over why it wasn't
> working, and I found that 2 directories within the pnfs namespace
> cause any process that attempts to access them to hang (ls, du etc)-
> all the others behave normally. These 2 directories are
> /pnfs/lancs.ac.uk/data/geant4 and /planck, both entirely unused. I've
> checked the pnfs placeholder files for anomalies but I haven't seen
> any. This is causing the GIP to hang halfway through getting the space
> information, and failing to publish. I could hack something together
> to get round this but I'd rather get to the bottom of this wierd pnfs
> behaviour. Any idea's on how I should proceed? I had a few ideas, but
> they fell flat because every attempt to access anything within these 2
> directories hangs.
>
> cheers,
> Matt
|