How to Perform a Active/Standby Switch on NCSS Real-Time Systems
Read the section on determining the real-time system status for more information.
- Log into either ucbrt or mnlort1 and examine the current eqassemble log file in /home/ncss/run/log/. Check the date of this file to see when it last changed. Then run “tail -f <eqassemble-log-file>” to see whether a new event is being assmbled. Only after you verify that no new events are in progress should you continue with the next step!
- Log into mnlort1 and ucbrt (using ssh) as user ncss. This can be done by one or two people.
- Run the command set_role active on the machine that is to become Active.
- Run the command set_role standby on the machine that is to become Standby.
- Check the status using the command get_role. This will report the current status of the machine as Active or Standby.
- Proceed to the procedure for switching post-processing roles.
The set_role command will send email and pager messages to report the change of status.
In the event that the machine which was Active crashes and is not available for login, you can proceed immediately to make the former Standby machine into Active. Whenever a real-time machine boots, it will set its status to Standby as part of the startup process. This will cause email and pager messages to be sent announcing that the booted machine is Standby.
After doing a Active-Standby switch of the real-time systems, it is important to switch the post-processing systems so they are running on the same side of the Bay as the Active real-time system, mentioned in the list above. Otherwise database replication problems are likely to happen.