I'm interested in one information that I cannot find anywhere.
Consider a simple 2 Nodes + Witness solution (Small Business/ROBO)
Consider this timeline:
1- Witness Fail : I have still 2 copies of data, so full data redundancy (and no witness), but ok with quorum
2- Secondary host fail: primary host have all the data uncorrupted, but have no quorum. All objects unaccesible
In this situation, is there a way, with or without opening a case to the support, to force VSAN to run with only 1 node and no quorum (like the first setup of VCenter on top of the same VSAN cluster, where you have only one node of VSAN and VCenter works)?
If yes, is it also possible with the secondary server up and the primary down?
Does the deployment of a NEW witness recover all my unaccessible objects?
No, there is no way from the GUI/CLI/other customer-facing method of making usable data from a single Stale replica of data - whether GSS can help recover some/all/any of this data to a usable state depends on too many factors outside of the scope of what I can discuss here and you should by no means be mistaking FTT=1 with 2 failures as anything less - as with any platform you should have good backups and/or DR solution regardless.
Deploying a new Witness would give you exactly just that, a new Witness with no data and not relevant to the quorum of Objects that already exist (and have no quorum and/or are stale).
- *In theory*, if the Witness is already gone then it shouldn't matter so much (from a data integrity and currency perspective) which site becomes inaccessible e.g. the relevance of Preferred and Secondary doesn't mean much when you don't have a Witness to care about this (other than potentially which site VMs are running when the 2nd failure occurs and how the failure occurs).
- This is not the only potential scenario, the more preferable (in theory) way for this to fail would be one site (again, doesn't matter much which) and then Witness - this is obviously preferable as we are left with a current single set of data.