Solaris

I'm using clustered zones on my machine. i'm only at the test phase of my design and ultimately the oracle zones will be using VxVM.

When the testing phase is complete, VxVM will be used in the containers. It is necessary for VxVM to run in the global zone for the containers to use it (is that true?).

I read (not on Sun docs) that SVM has to be installed regardless of VxVM being used, when using SC3.2, is true?

When using clustered zones how will the cluster deal with the VxVM filesystems when a zone fails over to another global zone server?

i can tell that SVM is allways part of a solaris installation since solaris 9. can't answer the rest of your questions...

I have a similar configuration setup, but we are using VCS instead of SC.

Yes, Veritas volume manager runs in the global zone

Code:

# zlogin -l erice0x fcprod
# vxprint -ht
VxVM vxprint ERROR V-5-1-11110  Please execute this operation in global zone

In what context does SC require SVM? That seems hard to believe.

We use cluster software, VCS, to mount and umount vxfs file systems into local zones the same way we mount and umount vxfs file systems in global zones. We simply specify the mount point as a directory inside the local zone.

Code:

From the global zone...
# df -hZ -F vxfs
Filesystem             size   used  avail capacity  Mounted on
/dev/vx/dsk/FCTIPROD_dg/FCTIPROD_vol
                        46G    10G    34G    23%    /zones/fctiprod/root/FCTIPROD
/dev/vx/dsk/HISTDW1P_dg/HISTDW1P_vol
                       100G    81G    18G    83%    /zones/histdw1p/root/HISTDW1P
/dev/vx/dsk/FCPROD_dg/FCPROD_vol
                       849G   701G   147G    83%    /zones/fcprod/root/FCPROD
/dev/vx/dsk/FCPROD_dg/backup_vol
                       749G   663G    81G    90%    /zones/fcprod/root/backup

From one of the local zones...
# zlogin fcprod
# df -h -F vxfs
Filesystem             size   used  avail capacity  Mounted on
/FCPROD                849G   701G   147G    83%    /FCPROD
/backup                749G   663G    81G    90%    /backup

you need some kind of volume managment if you use hastorage. so it is vxvm, svm or zfs. if you don't switch storage from one node to another you won't need volume management...

Firstly thanks for getting back to me, those commands have corroborated some of the assumptions i have been making.

I think the SVM package is installed anyway (by default), at this stage I dont think it's important I just wanted to verify for the benifit of my documentation.

what i have is:

Legacy database domains (on a 15K) we are migrating to branded zones, on two M5000's.

Those systems currently use VxVM. When they've been migrated to zones they'll still need the VxVM (on the global zone) to run. The containers themselves will be built on ZFS SAN. The oracle data groups are all mounted on UFS filesystems, because we all know; branded zones cant mount ZFS.

Sun cluster 3.2 using HA containers will be installed to the global zones.

What i need to know is: if a container fails over to the other global zone, what do i need to manually share with both of those servers, from a VxVM point of view, to enable that container to come back up?

I understand that SC will move the container over to the other global zone. Though, it wont be able to boot back up and mount it's filesystems if that global zone doesnt know about the others VxVM configuration, regarding the failed over container. So what do i need to do to ensure that happens?

VxVM won't share the SAN disks between servers, you will have to zone both servers at the switch. If you want to manually swing the disks after the disks are shared, simply vxdg deport, and then vxdg import the disks. I think you probably want to automate this process with the cluster software, and I believe the documentation you need is at this site below...

How to Register a Disk Group as a Device Group (Veritas Volume Manager) (Sun Cluster System Administration Guide for Solaris OS) - Sun Microsystems