Vsan ftt=1 and maintenance

[zdingelit]

Hello,

I want to clarify some things.

On a Vsan 6.2 streched cluster 8+8+1

Each Host runs 10 vm

As FTT=1 What is the expected behavior if hosts on preferred site are placed in maintenance mode one by one.

Thank you

[TheBobkin]

Hello zdingelit​,

If you place a host into Maintenance Mode with 'Ensure Data Accessibility' (EA) it will start moving all the VMs onto other hosts and then stop sending IO to the data-components on this node. Once this is done for all hosts on the Preferred site then you are basically running all VMs on the Secondary site off the remaining set of data on this site (e.g. FTT=0 until you bring the Preferred site back online and resync the data changes that occurred while it was absent).

Do ensure that there are adequate CPU and Memory resources on just one site to run all VMs comfortably without contention and/or consider which VMs will have to stay powered off (until both sites are up) if this is not possible.

If you mean just doing maintenance on one node at a time (with EA) then only a fraction of the data (~1/8th here) will be FTT=0 while this host is not participating - if you are only doing one node at a time then you can also consider doing maintenance with 'Evacuate All Data' (FDM) instead which will retain FTT=1 for all data during maintenance but take longer as data needs to be moved each time.

Bob

[zdingelit]

Thank you very Much for your quick and detailed answer. i thought that FTT=1 means we can shutdown one host, failed one host, fail one disk, fail one site, Lost the witness ?

[TheBobkin]

Hello zdingelit​,

"i thought that FTT=1 means we can shutdown one host, failed one host, fail one disk, fail one site, Lost the witness ?"

Yes it does mean this - but it achieves this by having a replica of all data at each site so if you have a failed site then the data is Failures To Tolerate=0 until you get that site back. Similarly if you have a failed host/disk the data that resided on this is FTT=0 until it is rebuilt elsewhere in that Fault Domain (site here).

While losing a Witness technically doesn't affect the data availability, this is required for quorum so this counts as a site failure and the data is essentially FTT=0 until it is restored/replaced.

Bob