(This is something you should be aware of, not a suggestion that we
all dive in with 1.4.0 today...)
Andrew
---------- Forwarded message ----------
Date: Mon, 9 Dec 2002 11:07:01 +0100 (CET)
From: Markus Schulz <[log in to unmask]>
To: [log in to unmask]
Cc: [log in to unmask], [log in to unmask],
[log in to unmask], [log in to unmask], [log in to unmask],
[log in to unmask]
Subject: [ITeam] UPGRADE TO EDGv1.4.0 <------ READ THIS MAIL
Dear Sysadmins and EDG Application Testbed-Users,
as you might know already EDG has released the version
v.1.4.0 on Saturday. Using the so called Fake-II (dbII) and including
quite some bug fixes from WP1 this will mean a step forward
toward stable operation (I said a step not a big leap forward).
At CERN we have already the RB used by CMS and one UI on 1.4.
(lxshare0383 and testbed013).
The upgrade to this version on the Application Testbed is planned to
be done today.
Since the current version and 1.4 are NOT backward compatible and
in fact using 1.4 RBs with older UIs will bring the system badly down
the sites have to switch the UIs and RBs at the same time.
We propose to start the upgrade at 12:00.
Here is now the sequence in which sites should do the upgrade:
0) Get the tag edg v1.4.0 from the CVS repository
1) Make your UI nodes unavailable for the users
2) Upgrade the UI,CE,.. (everything but the RB and FakeII)
3) Upgrade your RB (if you have one)
Switch off the services
Clean the RBs data base with the command:
pgsql -U postgres rb
delete from job;
"\q"
Upgrade your RB
Restart the services
This is required since the entries are not compatible and
bad things to your RBs future will happen if you mix them.
4) To test your new RB point it to the information index at
lxshare0225 at CERN. Submit a few test jobs.
The lines to change are in rb.conf in /opt/edg/etc
MDS_contact = "lxshare0225.cern.ch";
MDS_port = 2170;
MDS_timeout = 60;
MDS_gris_port = 2135;
MDS_basedn = "mds-vo-name=local,o=grid";
5) If your systems works send mail to the Integration team
list your status.
6) If your RB works with your upgraded UI you can let the
users in again. But it is very important that you avoid
mixing pre 1.4 and 1.4 versions of UIs and RBs!!!
Tell your users that they can only use the already upgraded RBs.
(If you are in doubt that you have sufficient control over
your users wait until all sites are at 1.4. before you let them
in again)
7) Those sites who operate RBs have to setup (on another node)
their own dbII. This can be done while the local RB is
running with the dbII on lxshare0225 at CERN.
You can find a simple HowTo at the end of the mail.
If your dbII is working switch your RB to it and announce to the
list that your dbII is ready.
8) We will announce when all the sites are upgraded.
Good luck
markus and emanuele
Here is the dbII HowTo:
1) Install the node with the standard BDII-cfg.h configuration file
2) In /opt/edg/etc copy the bdii.conf.template template file to bdii.conf:
> cp /opt/edg/etc/bdii.conf.template /opt/edg/etc/bdii.conf
3) Edit /opt/edg/etc/bdii.conf for your site. This usually means
changing only
BDII_HOST, BDII_PASSWD, and BDII_PASSWD_PLAIN.
To generate a new BDII_PASSWD(_PLAIN) pair (recommended!) you
can use the
/opt/openldap/sbin/slappasswd command which works just
like passwd for Unix.
If you are installing the BDII node for a testbed different from the
EDG
application testbed, then you also need to change MDS_HOST.
4) Copy the /opt/edg/etc/init.d/bdii file to /etc/rc.d/init.d (or create a
soft link), activate the service with chkconfig, and start the slapd:
> cp /opt/edg/etc/init.d/bdii /etc/rc.d/init.d/bdii
> /sbin/chkconfig bdii on
> /etc/rc.d/init.d/bdii start
*************************************************************************
* *
* CERN Markus W. Schulz *
* Bat. 31 2-015 *
* CH-1211 Geneva 23 *
* *
* Phone: +41 22 76 77909 *
* www.cern.ch *
* *
*************************************************************************
