Hi all,
My internal SAS disks (4 of them) are all sitting on the same controller/pcie device which is now being own by my default primary aka io aka control domain.
I have created a vdisk server, that serves slices on these 4 disks to guest domain and everything is working fine.
The issue is that, whenever I need to patch my control domain (reboot) the guest domain will be affected due to the fact that the entire IO/PCIE device is own by primary.
/SYS/MBU/SASHBA PCIE PCIE0 primary OCC
==================
q1) Does this means that I have no way to create another separate IO domain so that whenever I patch my control domain, my guests wont be affected ?
q2) If I have another 2 PCIE device (hba cards connected to SAN), can I assign them to another IO domain, then from that new IO domain, create a vds service and serve disks to guest domains instead ? (come to think of it .. this IO domain need its OS to run, unless I am doing SANBOOT, else this IO domain will still run from vdisk allocated from control)
q3) if a guest domain has no dependencies on the control domain, will patching the primary/control domain cause guests domain to go down ?
Regards,
Noob
Depends on the server, even rebooting of 'half' a server is possible.
If server has 2 or more sockets (T5-2, T7-2 .. ) you can have two (or more) service domains.
Lets say there are 2 sockets, with 2 controllers, 2 network cards 2 hba etc.
Those two service domains (primary and secondary) will hold 'PCI endpoints' to each of those mentioned above (1 per domain).
Guest domains (LDOMs) will be on on primary or secondary domain.
You can then actually achieve zero downtime when patching hypervisor.
Live migrating ldoms from primary to secondary, doing patching, reboot just primary etc.
But i would not advise it, if you don't actually require it real bad.
It increases complexity and maintenance.
If you have multiple sparc boxes you can always live migrate to another box.
Just be sure you are in 'the matrix' regarding OS versions and FW levels.
Also, read the fine print for OVM server, especially known issues 
Take a look at this article :
Availability Best Practices - Availability Using Multiple Service Domains (Virtually All The Time)
If you plan such configuration, plan carefully and read the exact docs for the version of Solaris you are using.
Hope that helps
Regards
Peasant.
1 Like
Hi Peasant,
Thank you so much for your reply!
The dilemma right now is that all the local SAS disk belong to 1 PCI endpoint and it is currently own by the primary/control domain.
The rest of the guests domains are actually having vdisks allocated out from these local SAS disks.
I can't allocate the PCI endpoint to any other IO domain as that would cause my primary to fail.
But everytime I patch the control domain, my guest will all have to be shutoff.
Is there anyway out of this ?
Regards,
Noob