How to restart Ganga job repository server

NOTICE: The Ganga job repository server is currently migrated from gangamd (lxgate31) to lxgate41. How to restart the new service can be found at the end of this page.

Status of the gangamd.cern.ch server in Lemon monitoring page: https://lemonweb.cern.ch/lemon-status/info.php?host=gangamd

Connect to the machine from lxplus:

$ ssh root@gangamd

AMGA requires PostgreSQL to be running on the same host.

Starting PostgreSQL

The PostgreSQL log is at: /var/log/postgresl

Checking if Postgres is running:

$ps uax | grep postgres

Start PostgreSQL:

$/etc/init.d/rhdb start

If PostgreSQL does not start, one possible cause is that there are locks from previous instances. To delete them:

$rm /var/lib/pgsql/data/postmaster.pid

AMGA

There are two versions of AMGA configured in gangamd, one with security running on port 8823 the other without running on port 8822.

Unsecure AMGA. Port 8822

The unsecure AMGA can be started as a service. Its log is at: /var/opt/glite/log/mdserver.log

After making sure that PostgreSQL is running, AMGA can be started with:

$ /etc/init.d/mdservice start

AMGA might also have left a lock from a previous instance. If AMGA fails to start due to problems with a lock, delete the following file:

/var/lock/subsys/mdserver

Secure AMGA. Port 8823

The secure version of AMGA is not configured as a service, and must therefore be started manually with the following command:

/opt/glite/bin/mdserver -c/root/mdserver_secure.config -D -l/var/opt/glite/log/mdserver_secure.log

The log is the file specified in the -l option.

New repository on lxgate41, port 8822

Status of the lxgate41.cern.ch server in Lemon monitoring page: https://lemonweb.cern.ch/lemon-status/info.php?host=lxgate41

Connect to the machine from lxplus (several GANGA developers are administrators on lxgate41, e.g. Kuba):

$ ssh root@lxgate41

The AMGA service is running on port 8822 and users must use SSL to connect to it. You can check whether it is still running by issueing

/etc/init.d/mdservice status

To be sure, check also the output of

ps aux|grep amga

you should see at least one process running like /opt/glite/bin/amgad -c/opt/glite/etc/amgad.config -D -l/var/opt/glite/log/amgad.log

If this is not the case, then have a look into the logfile (var/opt/glite/log/amgad.log).

You can restart amgad with

/etc/init.d/mdservice start

Have a look into the log file, whether there are problems with the database. The AMGA server uses an Oracle backend on int4r (later the LHCb RAC). In case you have seen connection problems, you can connect manually:

/usr/local/bin/sqlplus lhcb_ganga_r/XXXX@int4r

You can find the required password in opt/glite/etc/amgad.config, which is only readable for root, replace XXXX with it.

-- JakubMoscicki - 31 Jul 2006