Hi all,
Here's information about the origin of those "Populated file" errors
that some of you have seen when draining DPM. Jean-Philipe gives the
summary below, but there is other information further down. Of
particular interest will be:
svn ls https://svn.lal.in2p3.fr/Quattor/LAL/DPM/Tools
(I think the permissions have to be set correctly first of all)
Cheers,
Greig
-------- Original Message --------
Subject: Re: Populated files
Date: Thu, 21 Feb 2008 09:51:13 +0100
From: Jean-Philippe Baud <[log in to unmask]>
To: Greig Alan Cowan <[log in to unmask]>
CC: Michel Jouvin <[log in to unmask]>, dpm-users-forum
<[log in to unmask]>
References: <[log in to unmask]>
<9F19FE8C366C8F9B571B1803@[192.168.1.10]> <[log in to unmask]>
<[log in to unmask]> <[log in to unmask]>
There is a race condition which is triggered by the S2 tests. This race
condition can probably be also triggered by other applications in some
cases. I know that I've to fix this bug, but this one is not trivial to
fix. Sorry.
Jean-Philippe
On Thu, 21 Feb 2008, Greig Alan Cowan wrote:
> Ahh, so it really is being created. I was slightly confused since the file in
> question had a timestamp from a few weeks ago, so I was presuming that any put
> request would have completed long ago. Obviously this is the AbortRequest bug
> that you are talking about.
>
> Why are you so sure that this only happens with the S2 tests?
>
> Thanks,
> Greig
>
> On 21/02/08 08:34, Michel Jouvin wrote:
> > The message is 'file being populated'... and it means what it means! File
> > has been created in the namespace, there is a put request going on and not
> > yet completed. This is due to a bug in DPM handling of AbortRequest. You can
> > expect to have one such entry by run of S2 tests.
> >
> > Michel
> >
> > --On jeudi 21 février 2008 08:31 +0000 Greig Alan Cowan
> > <[log in to unmask]> wrote:
> >
> > > Hi Michel,
> > >
> > > Thanks for the information, but what does populated mean? Is it that the
> > > file is on disk but not in the namespace or something like that? Is it
> > > safe to completely purge them as you suggest below?
> > >
> > > Thanks,
> > > Greig
> > >
> > > On 21/02/08 00:36, Michel Jouvin wrote:
> > > > Greig,
> > > >
> > > > This is a know problem due to a bug in DPM when processing AbortRequest.
> > > > These entries are in fact the result of S2 test (test is correct!). I
> > > > have a small python script to purge those entries : just pass the name
> > > > (or list of names) of the replica to delete.
> > > >
> > > > To get it, the easiest is to use SVN and checkout the following URL :
> > > >
> > > > svn ls http://svn.lal.in2p3.fr/Quattor/LAL/DPM/Tools
> > > >
> > > > There is also a script drain.sh to ease running dpm-drain, taking care
> > > > of re-launching it when it exits with an error (or indefinitly if
> > > > --infinite) to workaround the DPNS time-out problem.
> > > >
> > > > Michel
> > > >
> > > > --On jeudi 21 février 2008 00:28 +0000 Greig Alan Cowan
> > > > <[log in to unmask]> wrote:
> > > >
> > > > > Hi all,
> > > > >
> > > > > Does anyone know what it means for a file to be in a "populated"
> > > > > state,
> > > > > according to the Cns_file_replicas table? I've seen dpm-drain
> > > > > returning
> > > > > errors corresponding to this bit of code:
> > > > >
> > > > > printf ("%s is being %s\n",
> > > > > lp->sfn,(lp->status == 'P') ? "populated" : "deleted");
> > > > >
> > > > > Thanks,
> > > > > Greig
> > > >
> > > >
> > > >
> > > > *************************************************************
> > > > * Michel Jouvin Email : [log in to unmask] *
> > > > * LAL / CNRS Tel : +33 1 64468932 *
> > > > * B.P. 34 Fax : +33 1 69079404 *
> > > > * 91898 Orsay Cedex *
> > > > * France *
> > > > *************************************************************
> > > >
> > > >
> >
> >
> >
> > *************************************************************
> > * Michel Jouvin Email : [log in to unmask] *
> > * LAL / CNRS Tel : +33 1 64468932 *
> > * B.P. 34 Fax : +33 1 69079404 *
> > * 91898 Orsay Cedex *
> > * France *
> > *************************************************************
> >
> >
>
--
|