Migrating to new HW

[JohnFjeldberg]

Hi,

we attempted to migrate two of the customers SG nodes from a rx7640 to a rx6600 (SG 11.19/HPUX11.23). We reignited the boxes onto the new hardware, represented storage and network (iostat to line up with old nic names). We had to back out as SG would not let the two nodes join the cluster as it was complaining about that the nic names (lan1 etc) did not match HW paths. Doing a string on the cmclconfig I see the HW paths are in there, but was not able to change them. It would not let me to reapply config either. Looked around but could not find anything, anyone knows if there is whitepaper/procedure how to migrate SG nodes to new hardware?

Thanks&Regards,

John Fjeldberg

[RAJD1]

John.

Perhaps you have to explore with a cutover technique,  with cutover technique it joins the node and you can start the packages on the new node.  During cutover of new node on new hardware, you need little downtime for stopping the package on old node and hald the old node, then join the new node on the cluster and start the package.  

Hth,

Raj.

[Stephen Doud]

SG  A.11.19 or newer will let you de-list specific LAN NICs from the cluster ASCII file and cmapplyconf while the cluster is running.  But you must remove the network device on each node that shares that subnet.  This will clear the old hardware path.  Then re-add the IP and NETWORK_INTERFACE entries and cmapplyconf again to add the current NIC hardware paths.

If all else fails, remove the cluster binary file on each node (/etc/cmcluster/cmclconfig), decluster the VGs (vgchange -c n <vgname>) and re-build the cluster and add the packages.

[likid0]

Hi,

I have been doing a lot of that latley, migrating clusters to new hardware. You should have no problem, what I normally do, if the hardware is inside the same datacenter(vlan/san), we move the nodes one by one:

Lets say we are moving node2 first:

Previous work:

Move al packages to node1.

Clone node2 OS to the new hardware(recreate mirror,etc).

Check al nics are ok on the new node2, and in the same vlans(linkloop with node1,etc)

Map all San disks, and import VGs

On the day:

Modify your cmclconfig.ascii to remove node2 from the cluster(you also need to modify you packages so they dont specifie node2 in them), you can do this modification without stoping the cluster.

After node2 is out of the cluster, do a shutdown of node2, and reboot your new hardware node2, with the real IP's, hostname,etc

Once the new hardware node2 is up, modify cmclconfig.ascii with the new nicks,cluster lock disk,etc for you new node2.

cmrunnode node2 to start it in the cluster.

Move all the packages from node2 to node1, so you can do the same procedure to move node1 to new hardware.

The only downtime you have is the package switchover between nodes.

Do you have the logs, and the all output of your procedure and also the errors you got, whan appliying the new config to the cluster?