VMware Cloud Community
vitaprimo
Enthusiast
Enthusiast

Turn off VSAN

I'm upgrading the infrastructure and I need to take the last host off from a cluster with VSAN enabled; basically it's a group of disks contained within a host, within a cluster with no other host in it. I thought about vMotion the VMs to other ESXi hosts but there aren't any online so I have to find a way of breaking the VSAN group apart without moving the VMs. Is this possible?

The any other way I could come up with is using storage vMotion only and use external storage instead, that way the VMs keep using the same hosts but are taken of the VSAN group and the downtime would be minimal, only the time the host takes to restart but I'm hoping if this can be avoided as I'd need to move all that data back. VSAN is still going to be breaked off, it's not just a restart, the host need to be taken off the uni-host cluster. There is a spare disk in the host unclaimed by VSAN but it's smaller than the required for all the VMs, I thought about moving part there, part off-host but I can't remember why is it there and why is it unclaimed so I'm a little uneasy about pushing its limits, also, I'm not sure if that'd let me take the host off the cluster, it's been a while since last time I've done this. Any advice?

I'll appreciate millions your advice, thanks! Smiley Happy

0 Kudos
4 Replies
TheBobkin
Champion
Champion

Hello vitaprimo​,

"I'm upgrading the infrastructure and I need to take the last host off from a cluster with VSAN enabled"

Can you elaborate on what these upgrades entail? e.g. will hosts be offline or rebooting multiple times for multiple hours and/or having hardware removed/replaced and if so, which components? If this is the last host, how did you proceed with upgrading the rest of the hosts?

"basically it's a group of disks contained within a host, within a cluster with no other host in it."

How do you have a one-node vSAN cluster? Is all the remaining data on vsandatastore as FTT=0?

"I thought about vMotion the VMs to other ESXi hosts but there aren't any online so I have to find a way of breaking the VSAN group apart without moving the VMs. Is this possible?"

What is currently running compute on this host and why are no other hosts available to vMotion these VMs to?(e.g. what happened to all the other cluster members?!)

Where is the data of these VMs stored - on the single remaining node in the vSAN cluster as FTT=0 data? Do you have hosts that are part of a vSAN cluster and/or have external storage (e.g. SAN/NAS) attached that you could XSvMotion the compute+storage of the VMs?

"VSAN is still going to be breaked off, it's not just a restart, the host need to be taken off the uni-host cluster. "

If you have completed remediating all the other hosts that were in the cluster previously, why can you not just enable vSAN on these, create a new vSAN cluster and vsandatastore, SvMotion the remaining data from the "uni-host cluster" to this new cluster, decommission vSAN on the remaining host once the data is off it, upgrade this host then add it back to the new cluster when complete?

If the above points indicate that I am misunderstanding your situation, please do elaborate.

Bob

0 Kudos
vitaprimo
Enthusiast
Enthusiast

I kept repeating the same to make in an attempt to clear things up but it backfired, I'm sorry; today I'm well rested though, what I wanted was to break off the VSAN volume and separate the its data to its discrete disks. Before giving up and going to sleep I left vCenter moving the data to central NFS stores; I didn't want to do this because of the traffic, disk wear and to avoid moving back the data--but it's all done.

Nevertheless thanks for your advise, and for the new terms I didn't know before, it's going to be my next weekend project. Smiley Happy

0 Kudos
TheBobkin
Champion
Champion

Hello vitaprimo

I re-read what you said and think I get what you intended.

Yes, it is possible to temporarily decommission vSAN on a node and add the disks back later, provided the vSAN disks are not touched (e.g. re-partitioned/formatted):

1. If this is the only node then power off all VMs (and unregister from inventory so no cleaning up of invalids).

2. Unmount the vSAN disk-groups (DO NOT use 'Remove', Remove=DELETE).

3. Note the UUID of the cluster (esxcli vsan cluster get) so you can add it back to this later and leave the cluster.

(note that 2. & 3. are likely extraneous as people have incidents with boot devices all the time and vSAN doesn't mind, but noting them here for clarity)

4. Re-install/upgrade the host then add it back to the vSAN cluster and re-mount the disk-groups.

VMware Knowledge Base

This raises the interesting question of where and/or if cmmds cluster info is persisted if there is only one host left in the cluster and it leaves.

Bob

0 Kudos
vitaprimo
Enthusiast
Enthusiast

I'm sorry about the delay. It's been hell the last months, I lost every VM thanks to bad Microsoft virtualization attempts of software. I have no idea how but when I started over vCenter was able to recover some clones and stuff I had laying around; stuff I do to test and never delete, that, that was never meant to be a backup or use for any other reason than to try out features or systems saved me. Well, that and iCloud, I had an emergency copy of the root's PFX, with only two clones and this tiny KB-sized file I was able to recreate everything, VMware is completely worth it, the really hard process this all takes was vastly shortened by vCenter cloning feature, Active Directory got really messed up so it all was still hard, but I could just clone another VM and try again. The other cloning "solution" still requires to set up and activate and all the checklist of getting a computer hand taking all the carton out, VMware doesn't it just works every time, as long as it is resource-happy.

Oh, vSAN is turned off now, BTW. It wasn't turned off they way I expected it to but it's off. Smiley Happy

0 Kudos