Holodeck vcf 9 - Cannot select datastore

[Lie_DW]

I installed the holodeck appliance and the firs time I tried to create an instance, the ESX vms were created so I got pretty far in the installation.

I try to install the lab in a 2 ESX cluster node. I created a SAN datastore of 2 TB. I tried both setting the targethost in the config to an esx and to vCenter.

I redid the installation and now I cannot get past the datastore selection. It keeps telling me I need to enter a valid option 

It only goes through when I select a local datastore but then it says my other esx cannot connect. Option 4 is a SAN datastore. The first time I did the installation it went past it but now I'm stuck at that point

[Dhruv Tyagi]

The target can either be a vCenter or stand-alone ESX not managed by vCenter. I think you already have that, but just to be sure.

Can you share more details on what is presented in the datastore list?

[Lie_DW]

When the target is my vcenter, It lists my san datastores and my local esx disks. 

I tried to detach the ESX so it's not managed by vCenter and when I create a config with the target host being my esx it even fails before the selection of the datastore. 

The weird thing is that the first attempt went through connection to my esx that was managed by my vcenter and with a standard switch and not a distributed switch but I do not thing the switch choice has an influence on the datastore choice tbh

[Dhruv Tyagi]

What is the error you see when you run the command after detaching your ESX from vCenter?

[Lie_DW]

Sorry I was out of office but here it is:

$error | select * gives this

[Hendrik Ruis]

Hi,
I have a similar issue but then with the network (next step).

After trying a VDS switch/portgroup gives my a extra failure on "cannot be NULL"

I was first trying with a host only but failed and tried erroraction silentcontinue but is was connected with a VC.

This give also other issues with the paravirtualized adapter on the nested ESXi host (which I had to manipulate manualy before booting)

So now I tried with a VC (and 2 host) > other issue because holorouter is on host1 and the deployment goes to host 2 (no VLANTRUNK for Nested Environment)
Removed host 2 > then continues correctly but now I got stuck on the No Trunk Port Groups.

Note : also DNS resolving on the holorouter deployment is heavily integred and holodeck command do not work if dns is not available.
Note2 : After shutting down the environment and start the holorouter / depot and vcf installer.. I get also nasty things of getting lost of configs (vcf installer does not recognize the depot settings > cannot be set later > error 502.

So after 4 new setups I am getting very frustated.
In the second setup I got as for as setup the vCenter in the nested enviroment but also failed > could not start observatory ...

[Hendrik Ruis]

5th time > pushed the new configuration to the esxi host directly (again) but failed on serveral items (random) but after set -loglevel debug  and pushed it 4 or 5 times the installation suddenly continues.. but stil with the paravirtual adapter issue. Now setup the basic managementonly config. 
Issue with vCenter also solved: https://knowledge.broadcom.com/external/article/377968/vcenter-server-upgrade-failed-due-to-obs.html

Still not seamless..

[Dhruv Tyagi]

Hi Hendrik,

The error in the screenshot you posted is because the powershell session was first connected to a vCenter and then you tried using a host as the target. To avoid this, just exit the powershell session and start from scratch with just the host as the target. Note the host needs to be stand-alone host and not a host managed by vCenter.

This give also other issues with the paravirtualized adapter on the nested ESXi host (which I had to manipulate manualy before booting) - Please provide exact error details so that we can understand better.

So now I tried with a VC (and 2 host) > other issue because holorouter is on host1 and the deployment goes to host 2 (no VLANTRUNK for Nested Environment) - I'm guessing you are trying to use a standard switch on vCenter? If yes, all hosts in the cluster need to have the same port group created. Otherwise, just use a vDS instead.

Removed host 2 > then continues correctly but now I got stuck on the No Trunk Port Groups. - Do you have a trunk port group on host 1? Is it a vSS or vDS port group? Is it configured to be trunk? The port group needs to be trunk and available on all the hosts within the cluster.

Note2 : After shutting down the environment and start the holorouter / depot and vcf installer.. I get also nasty things of getting lost of configs (vcf installer does not recognize the depot settings > cannot be set later > error 502. - Can you share more details, screenshots etc for us to understand this better?

In the second setup I got as for as setup the vCenter in the nested enviroment but also failed > could not start observatory ... - Can you share more logs of what you are seeing, this is not enough information for us to triage.

 

[Hendrik Ruis]

Finaly I have a successful deployment after around being busy for 7 hours installation of the nested environment.
ANd yes a target host managed by vcenter is not optimal but it worked for me after retries. 
Beside that I have a target DNS and NTP server that has to taken in account and need vCenter for other LAB's so this is very inconvenient.

Detail about the paravirtualized is shown below:

It failed to change it because the (targethost managedby )vCenter (192.168.10.199) seems to prohibit this.

Regarding the switches > yes I am using VSS > but on all host does not the trick because the VSS cannot have up-links > see documentation for preparations. VDS gives the same error > it just skips the question because it can't find any switch.
But after a 5th try it could.. 
What does a configuration of the VSS or PG makes it a TRUNK... so it should be detected by Holodeck?? > 4095 ?

The error 502 seems to be gone ... have now a successful connection > but this is only for deployment afterward you don't need it.

The vCenter issue is solved > second phase failed > had to set a timeout in the vCenter settings. But this seems to be persistent in the VCF deployment (this was the second time it happend)

After that the deployment of VCF went flawless.

Other findings:
It's seems to be impossible to remove other configurations I have noticed if the DNS record of the hosted platform is not available.
It got stuck on the remove-holodeck command. But will look into this, for now not a disaster anyway.

[Lie_DW]

Ok so I am able to go further in the process. This is what I did and now I'm stuck again :)

I have a vc with 2 hosts. I deployed the holorouter in vCenter on a distributed switch vlan.

I detached a host, staerted the terminal and powershell and I get to select a san datastore, not every datastore but I managed to select one and it's good enough for me. It deployed all the ESX vms but now I get this:

The holorouter is connected to a vlan that has connection towards the network and 2 ports that I created, the holo-pg A and B one like instructed. I can see that the ESX servers are connected to the holo-pg-a but there is still something wrong I guess

[Dhruv Tyagi]

Hi Lie_DW,

Are the VLANS 0,10-25 allowed on the physical switch interfaces connecting your ESX hosts? If not, either allow those VLANS or move holorouter to be on the same ESX host as your nested ESX VMs

[Hendrik Ruis]

Hi Dhruv,

Just for my clarification, when you look in the documentation (github) you can you either a ESX(i) only host or a vCenter Cluster.
This does conflict in your earlier statement that vCenter is, but the conflict is in using a ESXi host that is used / managed by a vCenter.

If you use a cluster environment there  you cannot not stretch the Trunk Port Group because they don't have adapters to a switch (install documentation) so how does the holorouter communicate with the deployed nested environment? It has to be on ONE host otherwise it wil not work on the other host.. or am I missing something?

[Lie_DW]

Hi,

Yes, I restarted and it was good but now onto the next one :D We're getting there, hopefully :D

[Hendrik Ruis]

Hi Lie_DW,

When I had this messages the nested hosted were not booting because of the controller was not a paravirtualized adapter.
The server does not boot anyway.
This only happens when a vcenter/esxi (v8) hosts are involved in the hosted platform I was using.

The vcenter seems to be getting in the way to change the adapter because it is not allowed by the vcenter when you used the ESXi host in the create new holodeck command in pwsh.
When the start failed I had to change the adapter manually an boot the server to complete the installation of the ESX host.

[Dhruv Tyagi]

Please note that ESX hosts managed by a vCenter are not valid targets for Holodeck deployment as mentioned in the documentation as vCenter does not allow certain operations on the hosts as you mentioned.

Either a vCenter or a stand-alone non-vCenter-managed host can be used as the target for Holodeck deployment

[Hendrik Ruis]

OK that is fine, but still if you have a a vCenter deployment (configuration) type for the 2 targets in the cluster;
If the holorouter is deployed on host 1 and the deployment of the nested hosted is done (automatically) on host 2 it failed on my watch.
The datastore I requested on host 1 was not accepted because of the host 2 was auto selected.

When I chose the datastore on host 2 (I dont have shared datastore > only local) the installation failed because it cannot find the Trunk PG's and if I create them there is no connectivity between them (the PG's cannot have adapters to the physical switch)
Result the nested hosted cannot connect to the holorouter functions.

So what am I missing here?