Make sure 'Data' check in vSAN is Health is green - check there is no resync.
Power-off all VMs running on the nodes.
Place all nodes in Maintenance-Mode (MM) with 'No Action' one at a time - this is generally the only time that MM with 'NA' is advisable (e.g. while running on back-up power during a cut, cut incoming or moving house).
Power them down from Web Client and connect via out-of-band-management to confirm they went down cold.
Wait until all nodes are back-up and connected to vCenter before taking them out of MM and take all out of MM within a few minutes.
Do you mean that your vCenter is not stored on the vsanDatastore or not registered on a node in this cluster? If not stored on vsanDatastore but running on one of these hosts then place this one in MM last from CLI or host-client and power off and on hosts from CLI or host-client.
Vcenter is not work on vsan cluster . vcenter is working another cluster. i have one vcenter with vsan cluster and another cluster so @i have one vsan cluster. Now I don't understand in this infrastructure . can I shut down or restart (reboot) any vsan ESXi host using the vSphere Client like right click then shut down or reboot?
vsans.JPG 13.6 K
Looks like you are using the C#/Thick Client - generally advise managing vSAN clusters via the Web Client(s) as you won't see the options for Maintenance-Mode options in the C# Client (it uses MM Ensure Accesibility by default).
"Now I don't understand in this infrastructure . can I shut down or restart (reboot) any vsan ESXi host using the vSphere Client like right click then shut down or reboot?"
What are you trying to achieve here?
As I said above, if you are planning on taking the entire cluster offline for a reason (e.g. moving/re-racking the servers) this would be the only case where you would need to use Maintenance-Mode 'No Action' option and power off all hosts in the vSAN cluster as you said above "now I want to power off all vsan esxi server".
If you are just doing maintenance or updates/upgrades of vSAN nodes you can do this without VM down-time by doing them one by one with Maintenance-Mode 'Ensure Accessibility' option - allowing the data to resync between or use 'Full Data Migration' option to ensure failures-to-tolerate is maintained during maintenance windows (but this takes longer as it has to move data off nodes).
Do always use a Maintenance-Mode though (and consistently), don't just right-click and power-off/reboot hosts - each host in your cluster has a chunk of your vsanDatastore storage on it, you need to ensure you are not going to cause unavailability.