VMware on AWS: HCX. Please, don’t change anyting!
shot story: Customer changed the name of a network that was extended by HCX. The VMs on VMC that use this network don’t reach On-prem VMs on same network. On HCX I can’t unextend this network because HCX and vcenter hadn’t the same name! yes, I solved this problem.
If you are using HCX don’t manually change any information about all the objects that are used by HCX!
Write it in your mind. Don’t change anything!
HCX use a own database e connect information about objects with vCenter database. If you do one action i put there something will go wrong:
- don’t change the name of the vCenter Objects advertised to HCX through Compute Profile and/or through NE appliances (VM, networks, datastores).
- don’t edit HCX fleet by adding the networks or every others information.
Those operations will create inconsistency to the HCX database and will create problems for the future upgrades.
So, as I said, customer changed the name of a extended network. I tried to unexend it from HCX manager but it doesn’t work. We hve to use “forced” option, network could be unexended from HCX manager (and it disappear from the list) but HCX fleet interface still exist on the vm! I can disablet it but the system want to use it with a new network extension.
Now I have to perform an upgrade to a new version of HCX and disaster happens!
HCX manager can’t update the fleet with a “broken leg”. It is not a good situation.
I read some articles about similar problem with datastore name. On this example HCX fleet was moved to a datastore that wasn’t on deployment resource.
HCX manager and fleets was updated On-prem and On-cloud except one.
I was on a maintening-window so I decided to follow the white rabbit and I did those actions:
- unextend all network that was managed by this fleet
- edit service mesh and remove 1 network extender
- wait for end of actions
- re-edit the service mesh and add a new network extender
- a new fleet wil be deployed
- re extended all network that I previously unextended
The configuration of this new fleet is correct and without some manual misconfiuration.
So I proceeded with the update. I tought I solved all the issues but after 20 minutes I saw a new VM with the name “<name-of-flee>-update”. Some process during HCX update was go bad and the manager recreate a new vm with same IP address of the new fleet. I immediatly shutted down the VM, I checked that all things work and I manually remove this VM.