Switching between the rtcpclientd and tapegatewayd daemons

WARNING

Please do not switch the tapegatewayd daemon on unless you are performing tests or you are running the CASTOR repack instance at CERN.

The tapegatewayd daemon is not to be used in production until it has been proven to work successfully with the CASTOR repack instance at CERN. This is still currently not the case. All experiment production instances and all external sites must use rtcpclientd until the official announcement is made.

How to switch from using the rtcpclientd daemon to the tapegatewayd daemon

1. Log in as root to the head-node running the rtcpclientd daemon
2. Copy the example retry-policy files /etc/castor/migratorRetryPolicy.pl.example and /etc/castor/recallerRetryPolicy.pl.example to /etc/castor/migratorRetryPolicy.pl and /etc/castor/recallerRetryPolicy.pl respectively. WARNING: Please dot not modify these files except for adding logging if necessary. These files currently configure the tapegatewayd daemon to retry both recalls and migrations for ever. This is the only safe option for the moment. Please do not write any retry polices that eventually fail recalls or migrations. Such retry-polices are not deemed to be safe.
3. Make sure you have a copy of the sqlplus switchover script switchToTapegatewayd.sql
4. Stop the following 5 daemons on ALL head-nodes (not just the head-node on which the rtcpclientd daemon is running):
   • mighunterd
   • rechandlerd
   • rtcpclientd
   • expertd
   • stagerd
5. Connect to the stager database using sqlplus and execute switchToTapegatewayd.sql
6. Start the following 3 daemons on ALL of the head-nodes where they were running:
   • mighunterd
   • rechandlerd
   • stagerd
7. Start the following daemon on the head-node that was running the rtcpclientd daemon:
   • tapegatewayd

How to switch from using the tapegatewayd daemon to the rtcpclientd daemon

1. Log in as root to the head-node running the rtcpclientd daemon
2. Ensure the migratorRetryPolicy.pl and recallerRetryPolicy.pl policy files of the /etc/castor/expert directory are up-to-date.
3. Make sure you have a copy of the sqlplus switchover script switchToRtcpclientd.sql
4. Stop the following 4 daemons on ALL head-nodes (not just the head-node on which the rtcpclientd daemon is running): :
   • mighunterd
   • rechandlerd
   • tapegatewayd
   • stagerd
5. Connect to the stager database using sqlplus and execute switchToRtcpclientd.sql
6. Start the following 4 daemons on ALL of the head-nodes where they were running:
   • mighunterd
   • rechandlerd
   • stagerd
   • expertd
7. Start the following daemon on the head-node that was running the tapegatewayd daemon:
   • rtcpclientd

-- StevenMurray - 19-Apr-2011