Details of 1st upgrade

The 1st step in the upgrade procedure - installing the local nameserver - can be done 'at risk' and does not require a downtime. The local nameserver will run on the DLF machine on each instance. This is a rolling update; each instance will be done separately.

• log into the DLF instance as root for the istance being upgraded
• Modify the CastorConfig table in the stager database using sqlplus (or your favourite GUI)

    UPDATE CastorConfig SET value = 'castorns.ads.rl.ac.uk' 
        WHERE class = 'stager' AND key = 'nsHost';
    COMMIT;

• Make sure the following entry in the /ect/castor/tnsname.ora file on the DLF machine:

    NAMESERVER =
    (DESCRIPTION =
        (ADDRESS = (PROTOCOL = TCP)(HOST = cdbc08-vip.gridpp.rl.ac.uk)(PORT = 1521))
        (ADDRESS = (PROTOCOL = TCP)(HOST = cdbc09-vip.gridpp.rl.ac.uk)(PORT = 1521))
        (ADDRESS = (PROTOCOL = TCP)(HOST = cdbc13-vip.gridpp.rl.ac.uk)(PORT = 1521))
        (ADDRESS = (PROTOCOL = TCP)(HOST = castor151-vip.gridpp.rl.ac.uk)(PORT = 1521))
        (ADDRESS = (PROTOCOL = TCP)(HOST = cdbc04-vip.gridpp.rl.ac.uk)(PORT = 1521))
        (LOAD_BALANCE = yes)
        (CONNECT_DATA =
          (SERVER = DEDICATED)
          (SERVICE_NAME = nameserver.gridpp.rl.ac.uk)
        )
      )

• Install the castor nameserver daemon castor-ns-server-2.1.7-27

    yum install castor-ns-server

• Edit the file /etc/castor/NSCONFIG with the acount details (see NSCONFIG on castor100 for details)
• Add a new entry to Cupv between the stager and the local nameserver:

   Cupvadd --user root --group root --src <stager> --tgt <dlf> --priv ADMIN

where stager and dlf should be the real host names (i.e. of the form ccsc01.ads.rl.ac.uk)

• Start the nameserver daemon and make sure it is running OK

    > service nsdaemon start
    > service nsdaemon status
    > tail -f /var/spool/ns/log

• Edit /etc/castor/castor.conf on the machine, modifying the CNS HOST entry to

   CNS  HOST  c<instance>dlf.ads.rl.ac.uk

where instance is the name of the instance being upgraded

• Restart the other castor daemons to use the local nameserver.

    > service dlfserver restart
    > service jobManager restart

• Log into the instance stager machine
• Edit /etc/castor/castor.conf on the machine, modifying the CNS HOST entry to

   CNS  HOST  c<instance>dlf.ads.rl.ac.uk

where instance is the name of the instance being upgraded

• Restart the castor daemons to use the local nameserver.

    > service rhserver restart
    > service stager restart
    > service rtcpclientd stop
    > service mighunter stop
    > service mighunter start
    > service rtcplientd start

Note: The rtcpclientd and mighunter should be restarted as shown; an incorrect restart could cause migrations problems.

• Log into the instance LSF machine
• Edit /etc/castor/castor.conf on the machine, modifying the CNS HOST entry to

   CNS  HOST  c<instance>dlf.ads.rl.ac.uk

where instance is the name of the instance being upgraded

• Restart the castor daemons to use the local nameserver.

    > service rmMasterDaemon restart
    > service lsfd restart

• Log into puppetdev.gridpp.rl.ac.uk
• Go to the diskserver/instance directory, e.g.

cd /var/lib/puppet/files/disk-server/lhcb

• For each service class edit the castor.conf files under <svcClass>/etc/castor for each service class and modify the CNS HOST entry

   CNS  HOST  c<instance>dlf.ads.rl.ac.uk

• Commit the change once all entries are updated
• Log into puppet master
• Push puppet changes out to disk servers

    puppet_svn_export disk-server

This completes the modifications required for the upgrade. The stagr database cleanup can also take place at this point as documented in the referring page.