Hello Maarten, hello everyone
Thanks for your help.
I did indeed boot the machines... several times.
Actually I even stop them _before_ making any important changes to
the profiles, as I rather check (a bit) the xml results before applying them.
So, as it concerns the SE, the installed vdt_globus rpms are
in version 1.2.0:
[sysunix@barentz sysunix]$ rpm -qa | grep vdt_globus
vdt_globus_jobmanager_pbs-VDT1.2.0rh7gcc3-1
vdt_globus_sdk-VDT1.2.0rh7gcc3-1
vdt_globus_info_essentials-VDT1.2.0rh7gcc3-1
vdt_globus_rm_server-VDT1.2.0rh7gcc3-1
vdt_globus_jobmanager_condor-VDT1.2.0rh7gcc3-1
vdt_globus_jobmanager_lsf-VDT1.2.0rh7gcc3-1
vdt_globus_essentials-VDT1.2.0rh7gcc3-1
vdt_globus_rm_essentials-VDT1.2.0rh7gcc3-1
vdt_globus_data_server-VDT1.2.0rh7gcc3-1
vdt_globus_rm_client-VDT1.2.0rh7gcc3-1
vdt_globus_info_client-VDT1.2.0rh7gcc3-1
vdt_globus_info_server-VDT1.2.0rh7gcc3-1
I must have missed something at reboot time.
Anyway, I'm gonna have a deeper check on all packages.
Concerning the CE, checking
http://grid-deployment.web.cern.ch/grid-deployment/download/RpmDir/external/index_LCG-2_3_0.html
I noticed that the the 2 incriminated packages
lcg-bdii-3.1.13-1.noarch.rpm and lcg-info-dynamic-pbs-1.0.3-1.noarch.rpm
are not listed in the rpm list!!!
I saw lcg-bdii-3.1.11-1.noarch.rpm and
lcg-info-dynamic-pbs-1.0.2-1.noarch.rpm instead, while
cg-bdii-3.1.13-1.noarch.rpm and lcg-info-dynamic-pbs-1.0.3-1.noarch.rpm
are included in the rpmcfg-2_3_0/ComputingElement-rpm.h
I suppose I expected the so much to be present that I didn't check their
version carefully enough, and just relied on the shell completion facility.
So finally I dowloaded them, created the headers and it went OK.
Concerning pbs packages that disappeared from the CE and the WNs,
I also finally got them to be installed by lcfg : there was NO include of
lrms-server-rpm.h and lrms-client-rpm.h in the packages list. I fixed it in
the LCG-2_2_0 way.
Same for the CA-rpms that disappeared from every host that hadn't the
updaterpms.localpkgs flag set : there's no include in the packages list I
downloaded. I suppose that will bother me a bit when the next ca_ rpms
upgrade will arise, but I suppose I will be able to handle it at that time :)
Finally, the non lhc VO addition was also buggy : site-cfh.h doesn't contain
#define SITE_CFG
#include CFGDIR/vos-cfh.h"
#undef SITE_CFG
so I added it just afet the #defines handling the SA_PATH_* for the lhc VOs
and that gave me the support of the non lhc VOs back.
So, that's all for now, I still must carry on with the final tests, but
hopefully,
IPSL-IPGP has *upgraded* to LCG-2_3_0
As it was quite painful I still wonder if I havent missed something when I
cvs-checkout'ed LCG-x_y_z : was z supposed to be 0 or was it something like
00, 01 or even 1 :-?
Cheers -you bet-
David
On Saturday 29 January 2005 02:06, [log in to unmask] wrote:
> On Fri, 28 Jan 2005, David WEISSENBACH wrote:
> > Good evening ROLLOUTers
> >
> > I've been trying -and trying- to upgrade IPSL-IPGP-LCG2 to LCG-2_3_0,
> > on RH7.3 with LCFG.
> >
> > Seems to be successfull on the UI, and maybe on SE where no packages _at
> > all_ were replaced nor added so I really won't be surprised if it finally
> > failed (I had no time yet to check the rpm lists for the SE for
> > confirmation).
>
> At least the vdt_globus rpms should have been upgraded to 1.2.0...
> Did you reboot the SE?
>
> > But on the CE and the WN it went really bad :
> >
> > First, I was warned that
> >
> > [WARNING] updaterpms: Couldn't find RPM header file for lcg-bdii-3.1.13-1
> > [WARNING] updaterpms: Couldn't find RPM header file for
> > lcg-info-dynamic-pbs-1.0.3-1
> >
> > that surprised me a bit because the files .lcg-bdii-3.1.13-1.noarch.rpm
> > and .lcg-info-dynamic-pbs-1.0.3-1 are present in the same dir as the rpms
> > themselves.
> > When I run
> > genhdfile-static-402 lcg-bdii-3.1.13-1.noarch.rpm
> > genhdfile-static-402 lcg-info-dynamic-pbs-1.0.3-1.noarch.rpm
> > I get no complains, but that doesn't change anything.
>
> Login on the CE and check this:
>
> ls -la
> /export/local/linux/7.3/RPMS/external/.lcg-bdii-3.1.13-1.noarch.rpm
>
> You could have some stale NFS mount problem.
> Have you tried rebooting the machines?
>
> > Anyway, I would feel very very happy if that was the only problem I
> > encountered.
> >
> > The fact is that the installation leaves the computing nodes without a
> > batch system! I requested for PBS in the profiles, but the upgrade just
> > *removes* the packages listed in pbs-(server|client)-rpm.h, despite many
> > efforts.
> >
> > So finally, my idea was to install them again (I used the old LCG-2_2_0
> > profiles, not to miss something behind, thus the title of this post), set
> > the updaterpms.localpkgs flag to yes (with updaterpms.localpkgs set to
> > cdb), and then try the 2_3 upgrade again.
> >
> > So this didn't fail that much, expect that without (at least ?)
> > lcg-info-dynamic-pbs, the advertising of the queues by the ldap tool is
> > very close to swiss cheese : full of holes.
> >
> > So maybe I still could force the install of this two packages by typing
> > the appropriate rpm commands (hopefully I just take care of 1 CE and 4
> > WNs), but finally, I took the wise (?) decision to seek for further
> > advice before actually doing it, as I also felt this situation should be
> > reported somehow.
> >
> > Thank you very much for having read this message up to here.
> >
> > Yours, kindly,
> >
> > David Weissenbach.
|