XenServer USB Pass-through
-
Come production, it's planned to have a dedicated hypervisor that will be running our backup target VM.
But still I'm uncertain of how I'd protect that Backup target VM. I could use Xen it's self or maybe even NAUBackup at the Hypervisor level, but this seems a bit extreme.
I really only care about the Backup Target, and my previously Backed up VMs. Creating a full weekly, using NAUBackup at 8TB would be painful to say the least.
-
Why not just use a NAS with an NFS share? Or are you trying to save as much as possible?
-
@johnhooks said:
Why not just use a NAS with an NFS share? Or are you trying to save as much as possible?
Well the original plan was two compute nodes, and a third compute node that we'd use as backup.
Which could still work, but it's a single backup target. So I'm trying to figure out the best way to have another backup system in place for that oh crap incident. We likely will still be using StorageCraft on the windows servers themselves, for a separate file level backup.
-
@DustinB3403 said:
@johnhooks said:
Why not just use a NAS with an NFS share? Or are you trying to save as much as possible?
Well the original plan was two compute nodes, and a third compute node that we'd use as backup.
Which could still work, but it's a single backup target. So I'm trying to figure out the best way to have another backup system in place for that oh crap incident. We likely will still be using StorageCraft on the windows servers themselves, for a separate file level backup.
Oh ok. You could either use the NAS for that and have XO mount the NFS share from the NAS. I'm sure a cheap synology would work as a fail safe. You could also use B2 storage and store the xva files there. Either temporarily store them on the XO vm or just have something copy the main backup to B2.
That would be easy since it's a small number of files, it would make it a lot easier to recover from B2.
-
I'm confused. Correct me if my thinking of your setup is wrong.
VM host 1 - 11 TB usable storage production workloads (never more than 1/2 storage used to allow for host failover)
VM host 2 - 11 TB usable storage production workloads (never more than 1/2 storage used to allow for host failover)
VM host 3 - 8 TB usable storage XO and backup serverIn this situation, how is the host 3 any different than a Unitrends box? Why would you be worried about all of the backups inside the giant VM that resides on host 3?
-
Well you're not thinking about it wrong.
But the two "production" hypervisors would have enough storage to run everything today + enough room for growth ~11TB each.
The 3rd Hypervisor would be the backup target host. Which if someone happens to that host, then I have no backups. So it's a matter of having multiple copies on different media's
-
@DustinB3403 said:
So it's a matter of having multiple copies on different media's
OK so that's just the 3-2-1 rule - 3 copies of data, 2 of which are local on two different devices, 1 offsite.
You currently have 2 - 2 copies of data on 2 different devices.
You add an off site and you're probably covered, assuming this design meets your requirements. -
I updated my post.
-
@DustinB3403 said:
Well you're not thinking about it wrong.
But the two "production" hypervisors would have enough storage to run everything today + enough room for growth ~11TB each.
Are you doing replication between these servers for failover purposes? If not, what is the recovery plan? Restore all of the VMs from backup to the free space on the still functioning VM host? How much downtime on the failed VM host do you allow (while trying to fix it) before you start that process?
And, if you aren't syncing the two production servers, do you really need enough storage to run all of the workload in the case of a VM host failure? I ask - because in an outage situation, do you need to restore everything from the down VM host ASAP, or is there a fewer number of critical VMs you can restore requiring less storage?
-
The plan is for the servers to be in a Xenpool for if we need to perform maintenance or whatever the issue is.
Likely we'll use HALizard and get that configured for failover, purposes that is still to be decided / tested. As my manager would say "we can't ever have down time..." well obviously some down time has to be acceptable, there is no such thing as 100% uptime.
So to reduce it as much as possible would be the goal 1 hour of downtime.
-
So it appears to be impossible to do USB Pass-though, without either a USB over IP solution, or a PCI USB Card.
Neither of which I have at the moment.
-
For Knowledge sake - is HALizard something akin to StarWind's vSAN module?
-
@Dashrender said:
For Knowledge sake - is HALizard something akin to StarWind's vSAN module?
No, HA-Lizard has no storage component. DRBD, which is in the Linux kernel, is like the StarWinds VSAN module. HA-Lizard is a script that automates the setup of DRBD, Xen HA and other features and adds its own fencing so that you can do effective two node HA setups.
