VSAN FTT question

[billdossett]

I am trying to get my head around VSAN FTT and how I am going to shutdown my cluster in the event of power fail.  I have been reading about FTT but afraid I am not quite grasping it yet.

Basically I have a 4 node cluster all nodes contribute to the VSAN.

I have a VM that talks to my UPS now and I am writing powercli to shut everything down cleanly...  no problems so far... but then I though I would get clever.  I have a fairly beefy UPS and it will run one of my hosts, the switch and all my internet related gear, router access point etc for at least up to 2 hours...   so I have my ups monitor system, and my domain controller set to should run on host 1.  My goal is to shutdown all other VMs including the VCSA shutdown 3 hosts and just leave the first host running with my domain controller/DNS and the UPS monitor running on it until we get down to 10 minutes left and then the UPS monitor will shut down the DC and the last host and the UPS monitor dies with the last host.  If power comes back on it will stay running like that until I phsically power up the other hosts - or may do WOL to wake them up, but haven't had a lot of luck with WOL as of yet.. .nvr mind... stay focused.

So my real question is the VSAN and what happens if I shut down 3 of the hosts and leave just the first host running - to the VMs running on the remaining host...  if I have FTT set to 1, and I shutdown 3 hosts, will I not have problems? 

I clearly need to do more reading, but I'm on overload today and it just isn't sinking in.  If anyone can kindly point me to any simplified docs regarding this I would appreciate it, or explain in a sentence or two...   thankyou.

Bill

[TheBobkin · Accepted solution]

Hello Bill,

I could probably do it in one but FTT in two lines is better:

A default FTT=1 Object (e.g. a vmdk) consists of 3 components: data component, replica of data component (as the FTM is R1), witness component (tiny and used to break quorum).

Majority of these components (2/3 here) needs to be available for the Object to be accessible, so in your case shutting down all but one node would not result in anything being accessible.

With vSAN clusters in general, the best action to take with outages is to shut down all VMs and then place all nodes in Maintenance Mode ('No Action' option) in as near a time-frame to each other as possible and bring the hosts back online all together in as short a time-frame possible too - so if your UPS is capable of keeping all four nodes and other equipment up for 10-20 minutes then this is the best course of action.

Bob

[billdossett]

Hey thanks, seems like the best policy.. I was only going to leave the one running so I would still have DNS/DHCP  and could still use the internet, but sounds like that isn't practical - there is DHCP and DNS on the router though, so probably best to do as you say, shut it all down together and then just switch to using the router DNS... actaully can probably get my router to be a secondary come to think of it and  not have to change any client configs... and that will run even longer off the battery - cool 🙂  thanks again!  and thanks for the explanation of FTT!