UNIX for Advanced & Expert Users

Hi all,

I am preparing for a worst case scenario. Say i have a production server A, should A fails ( for whatever reason), i want another server B to take over. How can i move everything from A to B? Assuming i have regular backup of A.

I've searched in the forums, and briefly came across ufsdump for solaris. What is the equivalent of the same command for linux and unix?

For Linux, you can look at dump. It is like ufsdump in Solaris and will provide you with similar options.
And what do you mean by unix? As far as I know, Solaris is a unix system. So is HP-UX, and AIX and .... there's a pretty long list.

hi blowtouch,

thanks for the advise.

When you must have failover, you need to have the data disks on a (probably raid) file server separate from the two boxes: A & B. Then you can failover to B by simply assuring B has an identical full set of programs with the same directory structure that A had. Or environment variables to compensate.

You do this by compiling all production pushes on both A & B at the same time. The code on B sits unused until needed. You have to keep rev-levels of software the same on both boxes as well.

On failover, users login to B and away you go. Obviously, B has to have mounted the data disks before you can failover.

Use a clustering technology. Something like SunCluster or VeritasCluster. They are however what you can call a more advanced topic in Unix, but it will definitly give you the failover from A ----> B with full application uptime.

If clustering isnt your thing and all you have is tapes, then just restore the tapes to the new server and make system adjustements.

Another thing you can do is have a VIP created in your environment. Have all applications point to the VIP. Make sure both servers are up and running. One taking load and the other not. Then have a job created to change the DNS entry of the VIP to point to server B should there be a failure on A.

There are a lot of ways of doing this, implementing it has to do with the comfort level and SLA's that the applications require.

I hope that helps.
-S