how to replace physical switch in vsan environment

[tonecho]

we are planning to upgrade our physical switch, but i can't find recomment steps. we have five hosts connected to two stack physical switch, vsan is configured on this five hosts. is there a way to replace two stack physical switch without losing data.

[viquarhcimca]

its all depend on your design . 

All ways make sure you have backup of your infra , VSAN cluster is in health state and enough space to rebuild the data if there is any host failure . 

possible impact will be if not designed correctly the cluster might get into network isolation . 

how is your vSAN cluster connected to the two physical stacks. if properly designed with fail over for vsan network .

for upgrading the switch you will have to check that procedure with the switch Vendor. 

[JeremeyWise]

Is the replacement process to retain nodes in current configuration and just replace switch

Ex:  10Gb TOR switch vendor foo  replacing same kind of 10Gb switch from vendor bar

or is this a rolling technology upgrade

Ex: 1Gb TOR switch vendor foo replacment with 10Gb switch.. .. so each esxi host has to be taken down in rolling upgrade and so during that process you will have to either bridge old to new L2 netowork... OR.. take down entire cluster at one time to replace in one outage cycle.

[TheBobkin]

@tonecho As with any such changes, I would advise starting with ensuring you are taking a current backup of anything important in the environment.

If you are switching out both physical switches then the inter-node communication will not be possible and thus you will have to schedule a downtime for VMs to do this - in vSAN 6.7 and later ESXi comes with a script for assisting with cleanly isolating all nodes from one another, I would advise following the steps in the below documentation (it is same process for 6.7, unsure of why only shows these steps in the 7.0 version), note the power-off (1.j) and power-on (2.a) steps don't need to be done in your case (unless there is something from the switch-side that requires it but I can't think of anything that would require this):

https://docs.vmware.com/en/VMware-vSphere/7.0/com.vmware.vsphere.vsan-monitoring.doc/GUID-31B4F958-3...

After the change is made, you can validate the vSAN network is functioning as expected even while the vsan-vmks are still untagged (from the script) using vmkping (at the correct MTU) between the vsan-vmk IPs on the nodes then proceed with the 'recover' part and next steps.