attempting to upgrade g series cluster from 4.0.302 to 4.0.401 and it keeps failing. When it fails @7% I get mount iso error. However more concerning is that it changes the dns on my vcsa to point to the PSC. I upgraded a different cluster with (same source and destination builds) and it went through however it too changed the DNS on the VCSA to point to the PSC, changed the PSC to point to the VC and changed all the hosts to point to the PSC. We have always used our corp DNS servers.
Anyone having experiencing this or similiar?
It seems to be normal behavior when upgrading VxRail with internal VCSA of version 4.0.x.
With Internal VCSA in 4.0.x environment, VC and PSC has dnsmasq service that serve dns server function.
VC/VxRail Manager/ESXi all refer to VCSA as DNS server and then dnsmasq service on VCSA forward it to external DNS server.
Only PSC has its own dnsmasq service, PSC refers to itself as DNS server ant its dnsmasq service forward query to external DNS as well.
When upgrading, PSC/VCSA will go offline temoporary. During offline, ESXi and VxM needs to resolve hostname so they change their dns server from VCSA to PSC.
This is normal behavior.
I agree with this assessment as it applies to VxRail 4.0.x. We have had no issues with this and the cluster was deployed initially with this configuration.
I have heard from a SE that vSphere 6.5 (VxRail 4.5.x) / vCSA 6.5 no longer allows a DNS server service to be run from it and it was complicating upgrades. At the time this seemed like the SE was not repeating information from a support document and it was just something "he heard", so take with a grain of salt. We have not yet upgraded to 4.5.x so I can't speak to the reality yet.