I found that poking into the SpaceManager database was useful - the srm logs dump out the SQL query that they use. I also had a weird situation where the SpaceManager thought some pools had hugely negative space when it did its sum, leading to there apparently being no free space for writing.
Derek
> -----Original Message-----
> From: GRIDPP2: Deployment and support of SRM and local storage
> management [mailto:[log in to unmask]]On Behalf Of Matt
> Doidge
> Sent: 21 January 2008 11:00
> To: [log in to unmask]
> Subject: Re: dCache 1.7 to 1.8 upgrade
>
>
> Heya squires,
>
> The upgrade to 1.8 isn't a very labour intensive one if you're not
> planning on implimenting the SpaceManager. Simply update the rpms,
> edit the dCacheSetup to your requirements (the thing to watch out for
> in my opinion is the new pool metadata system settings), then run the
> reconfigure scripts and restart it all. Our pools took an age to start
> up the first time as they built the metadata databases. The srm 2.2
> workshop notes (
> http://indico.cern.ch/conferenceDisplay.py?confId=21405 ) has all the
> information you'd need to do the upgrade, although scattered
> throughout a few presentations.
>
> Our troubles with 1.8 way back when we upgraded were all fixed by
> updating from 1.8.0-4 -> 1.8.0-5, Patch 5 fixed a bug or two that was
> plagueing us. Our current troubles- well I wish I knew exactly what
> was causing them! Even a complete rebuild of our PoolManager.conf
> hasn't fixed us, it refuses to believe we have any write pools.
> Something has broken our Pool Selection system. I would advise
> upgrading to 1.8.0-12 without activating the SpaceManager, then
> activate after you're comfortable that it's all working.
>
> cheers,
> Matt
> >
> > On 21/01/2008, Greig Alan Cowan <[log in to unmask]> wrote:
> > > Hi Chris,
> > >
> > > Latest dCache is 1.8.0-12 and I would recommend Java 1.6.
> > >
> > > For Matt, the upgrade to 1.8.0-12 hasn't gone so smoothly
> and he is in
> > > downtime at the moment. All services have come back up, but he is
> > > getting "no write pools configured" error message when
> trying to write
> > > to the new system (using the same PoolManager.conf as he
> previously
> > > had). We are looking into this.
> > >
> > > Note that you should run the DB migration script:
> > >
> > > http://trac.dcache.org/trac.cgi/wiki/SpaceManagerchemaMigration
> > >
> > > This should be done after the upgrade and running of
> install.sh, but
> > > prior to starting the SRM.
> > >
> > > Something you should note is that in dCacheSetup you need
> > >
> > >
> SpaceManagerLinkGroupAuthorizationFileName=/opt/d-cache/etc/Li
nkGroupAuthorization.conf
> > >
> > > Note that you should *not* have "" round the path name,
> even though the
> > > Book documentation states oterhwise.
> > >
> > > Also, in your utility.batch, you should comment out the stanza:
> > >
> > > create diskCacheV111.services.SpaceManager SpaceManager \
> > > "default \
> > > -export \
> > > -jdbcDriver=org.postgresql.Driver \
> > >
> > >
> -jdbcUrl=jdbc:postgresql://${spaceManagerDatabaseHost}/${srmDbName} \
> > > -dbUser=${srmDbUser} \
> > > -dbPass=${srmDbPassword} \
> > > "
> > >
> > > Since this refers to the old SpaceManager, not the new
> SrmSpaceManager
> > > (having both can lead to confusion on the administrators part).
> > >
> > > Let us know how you get on and if you need a hand with
> the configuration.
> > >
> > > Cheers,
> > > Greig
> > >
> > >
> > > On 21/01/08 10:23, Brew, CAJ (Chris) wrote:
> > > > Hi,
> > > >
> > > > I'm in downtime and just preparing to upgrade my dCache
> from 1.7 to 1.8
> > > > but I cannot find much in the way of documentation for
> the upgrade (at
> > > > least not official dCache pages).
> > > >
> > > > Do you have a good recipoe to follow?
> > > >
> > > > Also what version of Java do you recommend, I'm at 1.5
> but there are
> > > > recommendations on the NGDF pages to go to 1.6.
> > > >
> > > > Thanks,
> > > > Chris.
> > >
> >
>
