Hi Micheal,
We have found with our environment it depends on the number of cifs sessions you have connected to your filer/vfiler. The failover is delayed signficantly if you have hundreds of connections as the filer will try to be nice and try to notify each connection. This results in an extended period of outage. If you don't have hundreds of cifs sessions you can probably just execute the normal cf takeover/failback commands - probably need the -f if you have active sessions.
The upgrade advisor I have noticed when we have completed our upgrades also does really cater for vfilers. In our case we have a similiar setup in that the majority of cifs shares are presented from vfilers. You will need to execute: vfiler run * cifs terminate -t 0 to ensure it does all vfilers cifs services.
To answer you questions directly:
Is it enough to stop CIFS on the controller? Nope you need to do it on any vfiler with cifs sessions
Do not rather CIFS be stopped on every vFiler? y you need to use the vfiler run * cifs terminate -t 0
- How long CIFS is not available? it is short period of time with the terminate command and how quickly you type cf takeover
- After "cf takeover" only a few seconds until the other controller takes over? After some minutes when "cf giveback" was sent? Not sure what you mean with this question.
- How can this time be kept as short as possible? By doing this it minimises the actual outage due to cifs sessions
