How do I upgrade from one lcg version to another?

Even if nothing should happen during an upgrade something always can go wrong

0) Read the latest lcg2 installation guide that you can find at:

LCG Deployment page

1) Announce you are going to upgrade and an outage on the LCG-ROLLOUT list

LCG-ROLLOUT@LISTSERV.RL.AC.UK

2) Drain the queues switching off the mds on your CE.

3) Wait for the remaining jobs to finish (it will take some time but you can start with the nodes that frees themself before)

4) Download the new lcg-VERSION

export CVS_RSH=ssh
export CVSROOT=:ext:$CERN_USERNAME@lcgdeploy.cvs.cern.ch:/cvs/lcgdeploy
cvs co -d $NEW-LCG2-DIR lcg2

and check the differences. On the lcg2 installation notes normally the differences with the previous version are described quite well, it is advisable to read the carefully because there are few things that might need manual intervention and can't be done with the profiles. For the configuration files you can make a cvs diff on the examples directory to see what has been added or removed and apply the differences to your files, in particular site-cfg.h.

5) Replace the /etc/updaterep.conf with the new one.

cp $newlcg2-path/tools/updaterep.conf /etc/updaterep.conf

6) Do the following

cd $NEW-LCG2-DIR/tools
/usr/sbin/updaterep ./lcfgng_server_update.pl  $NEW-LCG2-DIR/rpmlist/lcfgng-common-rpm.h
/tmp/lcfgng_server_update_script.sh
./lcfgng_server_update.pl $NEW-LCG2-DIR/rpmlist/lcfgng-server-rpm.h
/tmp/lcfgng_server_update_script.sh

/tmp/lcfgng_server_update_script.sh

The last one needs to be run twice in case some dependencies failed.

Copy the new rpm lists from the $NEW-LCG2-DIR/rpmlist to your favourite location which by default is /export/local/linux/7.3/rpmcfg

There is no need to reinstall nginstallroot

Recompile your profiles and try on one node

rdxprof NODE-NAME

Check the new rpms have been picked up (in particular when there is a kernel upgrade).
In case of kernel upgrade it is better to reboot

reboot

GENERAL NOTES

If NFS is not working properly updaterpms doesn't pick up the new rpms. In case of reboot this is particularly annoying because while /boot/grub.conf has been updated to look for the new kernel the rpms haven't and the machine doesn't reboot unless totally reinstalled.
Reinstalling all the nodes and the CE to have a fresh installation is a way to be sure that everything is updated correctly.
Reinstalling the SE though requires also to the experiment software managers to reinstall their software. It is suggested to do it by hand.

2_0_0 -> 2_1_1 NOTES

Don't forget after upgrading the CE to make sure that the experiment specific runtime environment tags can still be published. For this:

mv /opt/edg/var/info/<VO-NAME>/<VO-NAME>.ldif files to <VO-NAME>.list

In the new updaterep.conf there is a missing directory called apps_common to add it:

edit /etc/updaterep.conf and add
MIRROR 73 7.3/RPMS/apps_common http://grid-deployment.web.cern.ch/grid-deployment/download/RpmDir/apps_common/\ index.html




Last modified Tue 10 August 2004 . View page history
Switch to HTTPS . Website Help . Print View . Built with GridSite 1.4.3
For more about GridPP please contact Neasan O'Neill