Vsan error on all server

[ralphte]

I am running in to a error when ever I try to write to my vsan.

"Failed to create object. Failed to place witnesses. There are currently 0 usuable disks and 1 more usable disks are needed in witness node."

I see this error showing up for all three servers when ever i run the proactive VM creation test.

I can write to the vsan data store.

I have tried rebooting everything and rebuilt the vsan datastore non of which has helped.

My health test shows everything is good.

After so more testing I found that if i took offline .17 I would get this

SO now i can write from two machines. If i bring .17 back online i cant write to any of the three.

Anyone have any ideas to what might be the issue. Thanks!

[thomashirsching]

The problem is that one or more ESX servers were members in a stretched Cluster.

The solution was to change an advanced settings back.

vsish -e set /vmkModules/vsanutil/stretchedClusterMode 0

[VictorQ]

Thank you thomashirschinger​

This command worked for us:

vsish -e set /vmkModules/vsanutil/stretchedClusterMode 0

[salkarkhi]

Victor, can you Please share with us when did u run the commend ? is it on the ESXI who having a problem with connection ?

[TheBobkin]

Hello salkarkhi,

The above is a very specific issue so do ensure that you are facing the exact same issue for the exact same reasons before running vsish set commands.

There are a number of scenarios where insufficient number and/or types of Fault Domains are accessible for component placement during Object/VM creation, which would cause VM creation or proactive test to fail.

The above examples appear to have had either a misconfigured stretched-cluster or an incorrectly decommissioned stretched-cluster - check whether this is even the problem (always run get commands before (and after) any set commands),

if you ARE running a stretched-cluster then this command run on hosts in that cluster should return '1' (e.g. true):

# vsish -e get /vmkModules/vsanutil/stretchedClusterMode

and if you are NOT running a stretched-cluster then this should of course return '0' (e.g. false).

In summary, it is easy to check whether this parameter is set properly and to rule out whether this is the issue you are having or not.

Hope this helps.

Bob

[karnakov]

Great thanks man! Very helpful after recreating vcsa with remaining esxi hosts config