Solaris - physical to virtual migration on the same M5000

Hello,

Firstly, apologies if the theme of this post is discussed elsewhere.

At the moment we have a dual-domain M5000 running. Each domain is running with equal amounts of CPU and memory.

What we'd like to do is move the 2 hosts in question (1 per domain) to a Solaris 10 zone of it's zone. Therefore, the same M5000 would be running 2 zones as opposed to 2 domains. All I know is that it's possible to go from physical to virtual with the use of a flar (although I am concerned about how Veritas SAN filesystems will be handled - boot disks are local and under SVM control) but cannot get my head around the symantecs of using the same M5000 to do so. I also gather that the domains of an M5000 can be "collapsed" to create a beast of a server but I've not done this as yet.

If anyone has any thoughts or ideas on how to achieve this, I'd very much appreciate it.

There are bound to be details I've missed which might be pertanent to figuring this out so will be happy to provide more info if I can.

Many thanks!

so u wanna migrate system from physical machine to virtual machine?
this way might work out: backup the physical system to an image, and convert the image to a virtual disk which can be directly used in virtual machine.
i know that easeus todo backup can do this.

Hello nm-longnumber.

It is easier than you think so .... Just do it :slight_smile:

What exactly is Veritas SAN filesystems? Are you using vxfs and/or vxvm ?

If your applications and data is on external disk then this will realy be quite easy.

Steps:
Create a flar for each domain. Google how-to-do-it

  1. make a text file of the file systems to exclude (everything on external disk, as well as the path to where you store the flar image).
  2. run the flarcreate command to create the archive.

shut down the second domain.
Add the systemboards to the ACL for the first domain.
add the boards to the domain.
restart the first domain.
check the hardware - you should see double the disks, cpus, memory, IO controllers. (cfgadm is your friend)
Import the vxfs file systems from the other host.
create a new data structure / file system layout for your zones.

This is where it becomes difficult. No solution seems perfect. Do you use snapshot file systems?
Do you have to have raw device access in the zones?
Do you use ZFS quotas?
Should you pool all the disks into a single pool and let ZFS manage space across everything in a single pool?

Unless you want to do something specific, steer clear of raw device allocations. Mount the file systems in the global one. use lofs to re-mount them into the zones.

Create your zones.
Do your resource management the way you want.
boot up the zones and test.

You can get fancy - minimize some downtime by not killing the second domain at the start. In stead you share the disks via NFS from the second domain to the first, create the zone for domain 2, stop the apps in the second domain, start them in the domain-2 zone, create the zone for domain-1, move the apps from global to zone, etc. It becomes a bit difficult only about deciding exactly where you rename the first domain to have a different name from the new zone which takes over its workload. Use sys-unconfig, reboot the domain and give the zone the name of the first domain (sys-unconfig the zone and reboot the zone). You move the IO controllers and CPU resources once you are sure that all is working, then re-configure the domain-2-zone to use the disks directly in stead of via NFS.

With more detail you can plan and perform this without issues. Just accept that you will access file systems indirectly. :slight_smile: