> 1. you need to have disksets and the SSA attached to both hosts (of
> course). They cannot share the disks at the same time!!!
That applies to SDS and the standard setup for VxVM only.
You can share physical disks with VxVM (i.e. you have more than one
diskgroup on a single physical disk) if you are willing to part with
the GUI and create the diskgroups yourself. Don't put any of the
disks you want to share into the rootdg diskgroup. My setup is as
follows: one VxVM disk in rootdg (simple, slice 7 on the internal
controller) for the vxconfigd to boot, all disks in the SSA have two
sliced VxVM disks that belong to diskgroups SSA-0 and SSA-1.
> | Ok, here is the situation. I need to know if there is any way to set up
> | 2 machines (we'll say Ultra 2's for right now) with an SSA RAID array so
> | that if one machine goes down, the other will take over the processes.
> | However, we do not want to invest in HA software. Is there any way to do
> | this where one machine could possibly NFS mount say the /home dir if
> | the primary machine were to ever go down?? Can filesystems be shared when
> | using SSA??
The SSA is a disk array, but not a RAID array per se. That is done by
software, either SDS or VxVM in that case. Automatic takeover is
possible, but rather inconvenient without the proper software. The
problem is not the scripts per se, but rather that you need to test
them very thoroughly. Failover is not good if it destroys data on the
way. Alternate Pathing and NFS client failover should help you a bit.
I'm currently looking into the same issue as you.
For fun I tried if it works to mount a filesystem to both hosts at the
same time for load sharing and it does, sort of. Firstly, let me make
it absolutely clear that if you write to the fs, your dead (on one
host or both, depending). So it can only work with ro filesystems,
which rules out /home. Secondly, neither SDS nor VxVM does support
that, so you're on your own if you fry something. Ididn't look deeper
into SDS because it doesn't allow to partition a disk between hosts,
but I think from the manual that guerilla sharing might work along the
same lines. For VxVM it works like this:
1. Import the diskgroup on the first host normally.
2. Import on the second host as temporary and forced.
3. Mount -ro on both hosts.
4. Umount.
5. Deport on second host and don't write back disk information.
6. Flush diskgroup information from first just in case.
You should have replicated disk information on _all_ disks in disk
groups that you want to share. That needs to be done manually when
creating the disk group, so the GUI is useless here. Otherwise the
two hosts may decide to put their information on different disks,
which makes for really interesting effects and consequently hosed file
systems.
I'm not certain what needs to be done if one or both hosts crash
during the dual-head mount. I believe that nothing needs to be done
if the second host or both crash, if the first one crashes I think it
would be wise to reimport without the temporary option. At least
you'd have to prevent the first host from mounting automatically.
BTW, has anyone tried to use SDS ufs logging on VxVM filesystems?
Achim Gratz.
--+<[ It's the small pleasures that make life so miserable. ]>+--
WWW: http://www.inf.tu-dresden.de/~ag7/{english/}
Phone: +49 351 463 - 8325
by