Quick and Dirty How to Rename/Re-IP an HP-UX Server w/o Rebooting?

[Alzhy]

Just wanted to get your thoughts on how you do this w/o a reboot. As I envision it:

1.) Plumb down current I/F.
2.) Plumb up with new IP/route,etc..
3.) Set new hostname (hostname)
4.) Restart salient services...

Or is there more to it?

This is essentialy to proceduralize our "Poor Man's Cluster" Strategy - wherein very very large environments can be failed over to another environment that will have the same OS and software revs. Since we are essentially on a SAN where "swaying" of disks and diskgroups can be pre-planned (ready presentations) or scripted SAN gyrations - we avoid the huge delay in reigniting the OS, fixing the SAN, etc.etc... So essentially:
1.) MOve over salient security files, fstab, CFG files, etc,etc,
2.) Rename destination environment - hostname and IP, services.
3.) Import SAN disk groups... mount
4.)Restart the Apps/Db/etc..

All within minutes....

[Olivier Masse]

Why don't you do like ServiceGuard does and manually add your IP address as a second IP address on the top of your fallback server when you need it (i.e. ifconfig lan0:1 192.168.1.100). You will not need to restart any service that might go haywire by changing the IP of the server.

We used this "poor man" trick on servers a few years ago to implement a manual HA, everything was done in a script (vgchange, mount filesystems, ifconfig lan0:1, etc) and it worked well.

[Alzhy]

Thanks Bud...

But we're "poor" and do not really have $$ for redundant hardware. Besides, this is not a banking app we're dealing her.

I just actually wanted to know if someone had come up with HOWTO/steps to rename/re-IP a server w/o reboots, etc.....

On the ServiceGuard Front -- Any word on when the new truly HA ServiceGuard is supposed to come out? It is suppsoed to have vMotion like features -- stateless APPS failover/migration.

I know AIX will have that feature very soon..

[Steven E. Protter]

Shalom,

Very little more.

Make your changed intened for next boot in /etc/rc.config.d/netconf

If the change is not permament skip this step and follows yours.

The applications however might get a little annoyed by this procedure.

You might want to try something different if the apps will tolerate it.

Create a floating IP address that is supported by /etc/hosts entries on both systems. Create a little monitor script to check if services are working on the active node and if not, try and pull the ip address to the passive node and follow the second steps 1-4.

Still either way the chances of success depends on the app's tolerance of steps 3-4 below.

SEP