dHCI initialization - common VM Network

[Labmaster]

I'm working with a customer to initialize a dHCI cluster.

VMWARE:

We created a Cluster in vCenter and attached to HPE DL380 Gen10 servers is ESXi and NCM.  We configured the management network, and two additional vswitches for iSCSI-A and iSCSI-B.  The standard "VM Network" portgroup that is always installed with VMware is on the same vSwitch as the Management port.

In the stack-setup wizard on the Nimble array, we are stuck at server validation, on a rule called "Same VM Network Rule"

Description - "Checks if same VM Network exists on all ESXi hosts in the cluster"

Message - "ESXi hosts do not have a common VM Network"

Can anyone share how the "VM Network" is supposed to be configured fir dHCI.?

A screenshot of vSwitches in a working dHCI solution would be greatly apprecieated.

Thanks,

Kevin

[Sheldon Smith]

Search for "dHCI" at HPE InfoSight > Resources > Nimble Storage Documentation.

Check out the HPE Nimble Storage dHCI and VMware vSphere Deployment Guides

[VijayK-NimDCS]

Hi Kevin

The Rule checks for a Virtual Machine portgroup with the same name on all the hosts.  The standard portgroup called "VM Network" created during ESXi install  (that you mentioned) will suffice. 

Thanks

Vijay

[Nick_Dyer]

If these are new hosts (what we refer to as a greenfield solution) then there’s no requirement to configure any vswitches in vSphere whatsoever, nor build a cluster or add anything to vCenter - the stack setup will do that work for you.

Also, recommend to check out the Welcome Center - which details the whole installation procedure in detail with accompanying videos.
It’s great resource to have when doing installs of Nimble or dHCI. https://InfoSight.hpe.com/welcomecenter

[Labmaster]

We tried the "Greenfield" method, but server validation fails because ESXi is on an SD card.  The message says that ESXi can't be running on local datastores backed by SD card or USB.  These hosts were ordered without disks, we have run ESXi on SD cards for 10 years.

When we used the "Brownfield" method (create cluster add the hosts, then select that cluster in the wizard), the "ESXi local datastore rule" is just a warning, not an error.

In Brownfield, our last error is the "Same VM Network Rule".

[Labmaster]

We have that default "VM Network" port group.  Must be another piece of configuration that we are missing?

It would be great if anyone could provide details on what is actually happening in that check.

It makes sense to me that the data network (iSCSI) is checked with vmkping from the hosts to the storage array.  However, I have no idea how the wizard can check for a "common vm network" (other than checking that the same port group exists on both servers).  There is no vmk attached to the port group to use vmkping.

Feels like magic, and I want to understand the magic.

[Labmaster]

I have read all the documentation I can find, including the links you sent.  Nothing specific about installing ESXi on disks instead of SD cards, nor anything specific about the "COmmon VM network".

Documentation is still lacking specifics.

Supposedly 57 rule checks in server validation.  Those need to be outlined with specifics.

[VijayK-NimDCS]

Your observations are correct.

I am able to trigger/fail that rule when the port group has different names on the 2 hosts I tried. One was named VM network and the other VM network 1.  Renaming both to the same (does not have to be VM Network) resulted in a PASS on that rule.

Heres what I have on my hosts

Host 1
esxcli network vswitch standard portgroup list
Name                Virtual Switch  Active Clients  VLAN ID
------------------  --------------  --------------  -------
Management Network  vSwitch0                     1        0
VM Network          vSwitch0                     0        0
iSCSI Network 1     iSCSI1                       1        0
iSCSI Network 2     iSCSI2                       1        0

Host 2
esxcli network vswitch standard portgroup list
Name                Virtual Switch  Active Clients  VLAN ID
------------------  --------------  --------------  -------
Management Network  vSwitch0                     1        0
VM Network          vSwitch0                     0        0
iSCSI Network 1     iSCSI1                       1        0
iSCSI Network 2     iSCSI2                       1        0

[Labmaster]

That example helps a ton, thank you.  By "Active Clients" on the VM Network, does that mean you have 0 VMs?  We don't have any storage attached yet, so I can't make a VM (the whole point is to create storage from Nimble later in the wizard).

If you could check one more test for me, it would be appreciated.  What happens if your VLAN ID is different between the VM Network and the Management Network?

Normally, our management and VM use different VLAN.  I tested a couple of minutes ago making them the same VLAN, and it still failed.  I also created a new VM portgroup, as simple as possible "VM" making sure they matched, and it still fails.

[Nick_Dyer]

"We tried the "Greenfield" method, but server validation fails because ESXi is on an SD card.  The message says that ESXi can't be running on local datastores backed by SD card or USB.  These hosts were ordered without disks, we have run ESXi on SD cards for 10 years."

That's correct - Greenfield mandates the requirement for local drives in hardware RAID 1 - it doesn't support USB. In fact, if the servers were ordered via the Greenfield method then this would've been mandated in the quote tools and it wouldn't be possible to order USBs for local storage.

Don't forget you can call HPE NImble Support and get real-time help to resolve this problem, rather than having to rely on intermittent forum posts