Have you thought about looking at keepalived or linux-ha with perhaps linux virtual servers or some other IP swapping technology?
The basic idea of all high availability stuff is: - Some decision making to decide who should be master - Some fencing to ensure that the master is master (all kinds of fun acronyms here, like STONITH - shoot the other node in the head...) - IP aliasing or other routing technology to ensure packets flow to the master Obviously you can script all this yourself (which is basically what you are sketching out in your previous emails). But it may be worth looking at some of the well debugged projects so see if they solve any problems first? I'm unfortunately not talking from experience, but from a little knowledge I would suggest you look at keepalived and linux virtual servers. The basic idea is that you have a monitoring demon on each node watching the other (keepalived) and the demon makes various decisions on who should be master (other node dead, etc). When you want to be master you simply add the master IP to your network device and start answering queries (LVS)... The master IP basically flip flops between the nodes according to your rules (what should happen when the main node goes down, comes up, etc) For completeness you want some way to fence each node so that simple cables falling out can't cause wierd issues. However, you can probably not worry too much for your requirements Good luck - please share what you come up with! Ed W
