Brew, CAJ (Chris) wrote:
> Hi,
>
> I'm in the process of decommisioning a server so am trying to empty the
> pool on it.
>
> I've used the copy module to copy out each VO storage class to the VO
> pool group and then deleted all the files in the pool that have more
> than 1 replica listed in the companion database (good notes from Greig
> Cowen at http://www.gridpp.ac.uk/wiki/Draining_a_dCache_pool).
>
> Now I'm left with a about a thousand files which errored when
> copy-module tried to copy them.
>
> These all appear to be orphans, i.e the exist on the filesystem but not
> in pnfs.
>
> So I can see the file in the pool data directory, rep ls in the pool
> admin interface sees them and cacheinfoof <pnfsid> in the PnfsManager
> says they are in to pool but storageinfoof <pnfsid> reports
> "storageinfoof failed : Not a valid PnfsId <pnfsid>" and pathfinder
> <pnfsid> give " File not found : <pnfsid>"
>
> Is there any way to recover or do something sensible with these files -
> without some sort of path information, I doubt the VO will be able to do
> anything with them.
if there isn't a pnfs path for the files then the VO has probably
deleted it and for some reason the deletion hasn't been sent to the pool.
On your head node, does the /opt/pnfsdb/pnfs/trash/2 directory have
files with the names of the pnfsids?
Or does/opt/pnfsdb/pnfs/trash/2/current directory contain a file called
failed.<poolname>, does that file contain the orphaned pnfsids?
Derek
|