Migrating from NW 6.5 SP8 to OES2 SP2

Question

I'm looking to move from NW 6.5 sp8 (no clustering) to Oes2 sp2 (clustering). The main goal in this migration is to no effect on the user experience. The big issue is server name, I want to keep the existing server names. A lot of users have shortcuts that use UNC (\\servername\vol\filename). I looked at using an alias in DNS but the shortcuts still won't work.
So I'm looking at doing a migration with clustering all at one time. Of course with a clustering resource I can use the existing server name. The issue that comes up is when I'm doing the cut over, how will clustering react to creating that cluster resource with the same server name that already has a server object with that name. I brought this up to Xang(spelling) at Brainshare this year. I wanted to see if anyone has gone through this. I have setup a test network to go through the scenario and see what happens. Does anyone have any ideas on this scenario or been through this? Thanks in advance.

Answer 1

Originally Posted by timothyaw
I'm looking to move from NW 6.5 sp8 (no clustering) to Oes2 sp2 (clustering). The main goal in this migration is to no effect on the user experience. The big issue is server name, I want to keep the existing server names. A lot of users have shortcuts that use UNC (\\servername\vol\filename). I looked at using an alias in DNS but the shortcuts still won't work.
So I'm looking at doing a migration with clustering all at one time. Of course with a clustering resource I can use the existing server name. The issue that comes up is when I'm doing the cut over, how will clustering react to creating that cluster resource with the same server name that already has a server object with that name. I brought this up to Xang(spelling) at Brainshare this year. I wanted to see if anyone has gone through this. I have setup a test network to go through the scenario and see what happens. Does anyone have any ideas on this scenario or been through this? Thanks in advance.
From what I'm reading clustering is more a means to a migration question than wanting it as solution...?
1) You are going to be running in a virtual environment? (can have implications for clustering)
2) From what I read, clustering is not something you want for what it offers in HA terms... but more as a means to reach your goal of keeping the old server name?
Have you had a look at the Transfer ID option? It sounds like a good option in your case.
Another possibility that also keeps downtime to a minimum (have used this many times to split up migration tasks) - quick overview:
1) Setup or use an existing OES2 server to host the new data volumes. This server should be using storage that can easily be swapped over to the new OES2 server that will be replacing the NetWare server.
2) Create the NSS pools/volumes with the correct names, sizes etc
3) Use the miggui (or SCMT) tool to do the file migration. This will sync keeping all file attributes and metadata - this fase can be done a day or two before breaking down (removing from tree) your NetWare server.
4) Setup a new SLES 10 server with the OES2 media added in (but no products selected or configured!) so it's ready to be installed into the tree. At this time it has a temporary name and ip...
5) Decommission the NSS volumes from the temporary assigned server (offline NSS pools and remove the eDirectory pool and volume objects) and disconnect the disks from the server (so they are ready to be connected to the new server)
6) At the planned time remove the NetWare server from the tree using nwconfig, remove leftover objects for that server from the eDirectory.. then give it a new ip, change the name in the autoexec and reboot.
After reboot you can either install the server back into eDir in the existing tree or a temporary one (I prefer the last to keep it fully separate)
7) After having removed and renamed the NetWare server, check eDir health, reconfigure the SLES server with the correct name and ip and install & configure eDirectory using YaST's OES install and configure (also add nss).
8) When this is done and running ok, bring down the new server, connect the data disks to it and boot it back up.
9) Commission the disks to the server using nssmu <ALT><F4> on pools and volumes
10) Run a tool to sync changes from the old NetWare server to the new OES2 server (mirror sync, so the new data volume is identical to the old one).
I still use the old robocopy script to do this mirror sync as it works well.
DO TAKE CARE : exclude syncing the ._NETWARE folder on the root of each NCP volume... as you will otherwise wipe the trustee xml files.
To do a last check comparison, tools like beyond compare/Total commander can do a good job here.
Post jobs could include re-linking user home directories and default server assignments sing the HOMES tool (HBware).
If the NetWare server was also doing other tasks... you can choose to reconfigure those takes before doing the steps mentioned above, or reconfigure after. It's dependant on what you have running on the original server.
Cheers,
Willem