Print

Print


Overall the procedure to install SLC3 is reasonably well documented
and the example config files are very helpful. And overall the
procedure was remarkably painless, especially thanks to Steve
Traylen.

We tried to follow the install instructions as given in the documentation.
They did not work for us. So we did the procedure using the yum install
route. The yum install route was given to us by Steve Traylen.
It would have been quicker if this option had been
documented as an alternative in the install and configuration
instructions.

We had three problems which could be classed as bugs.
. After realizing that the INSTALL_DATE variable has to be set (the
example site config file did have not entry) it turned out that
the resulting system config file had a blank space after the
date which must have been introduced by the installation procedure.

Although the date has been entered without an offending blank space in the
input site-info.def config file ( I checked that using the "list"
environment under  vi ) each configuration will again introduce a blank
space in the resulting system file, which then has to be hacked by hand
to remove it.

. The installation procedure identified 3 processors per node in the
automatic configuration of the pbs (torque) system. So it
was necessary to change the file /var/spool/pbs/server_priv/nodes to
replace the default np=3 by np=2.

. On the monitorbox the tomcat process(es) started to respawn without
provocation and led finally to the system being totally unusuable and
needed to be hard reset to reboot. I sent a message to LCG-ROLLOUT
describing the symtoms. Steve Fisher told me that there was a new
startup script for tomcat which could be picked up from

http://hepunx.rl.ac.uk/egee/jra1-uk/LCG/current-release.html

After installing that and restarting tomcat with it, the system is
behaving normally. As the new script was available, the problem must
have been known, but this new up-to-date script is not yet part of
the rollout release. It would have been helpful if an alert had been
added to the installation instructions.

Another pitfall are the /etc/ssh/ssh_known_hosts files which go out of
date on a re-configuration. It would be nice to have as part of the
installation instruction an alert together with the command of how
to recover from that.

I hope that this will be of some help to others.

Lydia


--------------------------------------------------------
Dr E L  Heck
Department of Physics  | Tel.: + 44 191 - 334 3628
University of Durham   | Fax.: + 44 191 - 334 3645
Durham, DH1 3LE        |
U.K.                   | e-mail: [log in to unmask]
--------------------------------------------------------