Server migration in MCSG environment

[Bill Ye]

We have MCSG 11.14 and SGeSAP 3.07 running on two HP-UX 11.0 servers. Recently, we want to replace these two servers with newer hardware model running HP-UX 11i, and keep the same MCSG and SGeSAP version.

Because of limited downtime, I plan to migrate one server in one smaller downtime at a time. Is this solution doable? Anything should I prepare and be careful? Any suggestion?

Thanks
Bill

[melvyn burnard]

well my suggestion would be to keep the current cluster running, add in the new nodes to create a four node cluster, ensure all the work is done to allow the vg/lv/fs activation, and setup your package(s), test they will start and halt correctly on the new servers, and once you are happy, move the packages to the new nodes, and then remove the old nodes from the cluster, reducing it back down to a two node cluster again.

[Sanjiv Sharma_1]

Hi Bill,

You have two options:

If the storage is not in a SAN environment and cannot be shared among four nodes at a time then Create a seperate two node cluster of the new servers and migrate from old systems to new sytems. This will required data migration too.

Else, if possible create a four node cluster and slowly migrate one package at a time. This method is more convinient and safe.

[Bill Ye]

Thanks for your suggestions. More clarification to this migration.

1) The current two-node MCSG using direct connections to XP512.
2) There is no free XP port left for new servers, and new servers have to use Fibre cable from old servers.
3) And we want to keep the same hostname & IPs on new servers.

Thanks.

[Rajeev Shukla]

This setup will really take downtime as you cant see the devices on the new server you built.
The best i suggest, which will help you in future too is invest some money in buying 2 switches and connect your XP via a switch, that will give you true failover and expandability for future say to add a library or any other devices in your SAN

[Sanjay_6 · Accepted solution]

Hi Bill,

If you want to reduce downtime, you'll have to do it like this.

Convert the two node cluster to a single node cluster, remove the second server and install the new server in its place. Reconfigure the cluster as a two node cluster. Again reconfig the cluster as a single node cluster with only the new server in the cluster, add the new server and reconfig as a two node cluster.

You have to be careful since doing all this will involve a very careful planning and keeping track of changes and syncing up the configuration files across the nodes will have to be done very carefully.

If you are able to get identical device files for the disk and lan ports that may reduce a few things that you'll have to keep track of.

Keep a good copy of backup, maybe two to allow you to backout if there is any problem.

Hope this helps.

Regds