VMware Cloud Community
Rylos
Contributor
Contributor
Jump to solution

2 Node Cluster with witness. Can not finish configuration, step 5 is grayed out.

I have the following setup and I am trying to finalize it with vSAN so I can start load testing the setup.

-Two identical servers running 6.7.0 build 10764712

--each contains

---1.64TB spinning disk

---2 x 500GB nvme SSD. I didn't realize I couldn't use two separate cache drives unless I have more than one spinning disk. Lesson learned, can't raid, oh well.

---80GB DDR4

---2 x E5-2687W v3

---Intel two port X540 adapter - one port setup on maint/vmguest network and one port on vSAN iSCSI network

---built in 1Gbit NIC setup on maint/vmguest network

-Witness appliance running 6.7.0 Update 1 (Build 10302608)

--set to small so it has 10GB ssd and 15GB spinner

--two ports, one on maint/vmguest and one on iSCSI

Hosts and witness can talk to each other on all subnets. DNS resolves the maint/vmguest network ip. Witness is running on an entirely different ESX server and is not a member of the cluster, only the datacenter under vCenter.

I am selecting the "Two host vSAN cluster" from the wizard. Selecting my spinning disks and one ssd from each server(It actually auto picks). I get to the very last step and I can not finish. It's all green but it just says "review settings before completion", but nothing is on the screen otherwise. "Next" is grayed out. Tried this in chrome and IE11, same thing. Edge, can't even get past step 2 as the next button is gone completely. Firefox, same thing, no next button on step 2.

Any ideas here?

0 Kudos
2 Solutions

Accepted Solutions
TheBobkin
Champion
Champion
Jump to solution

Hello Rylos

Welcome to Communities and vSAN.

"---2 x 500GB nvme SSD. I didn't realize I couldn't use two separate cache drives unless I have more than one spinning disk. Lesson learned, can't raid, oh well."

Only one cache-tier device per Disk-Group - you could however go with less capacity and better performance and use NVMe devices for both capacity-tier and cache-tier - either way though if this is hardware that is not on the HCL and/or nested then don't be expecting the world with regard to performance and/or potential reliability.

"-Two identical servers running 6.7.0 build 10764712"

"-Witness appliance running 6.7.0 Update 1 (Build 10302608)"

Update your Witness to the same build as your hosts - just because the Witness is an appliance doesn't mean it should be neglected as it is a vital part of the cluster what with being its own Fault Domain.

You never mentioned what build your vCenter is on, if it is 6.7 pre-U1 then upgrade before attempting anything - while 'ESXi' 6.7 U1 + vCenter 6.7 GA are a supported configuration, vSAN 6.7 U1 and vCenter 6.7 GA are not as too much changed and various issues can be encountered even during attempts at initial configuration.

Step 5 as per the 6.7 UI client is claiming the Witness Appliance disks, do you see literally *nothing* or do you see a screen with no disks available for claiming?

You also failed to mention how the Witness has vsan/witness traffic configured, but then again Step 4 in the UI configuration wizard should prompt and not allow selection as cluster Witness if no vmk is enabled for vsan traffic.

"Hosts and witness can talk to each other on all subnets."

The proper way of checking this requires you validate they can communicate over the required interfaces at the configured MTU and on the required port

# vmkping -I vmkX <Dest-IP> -s 1472 -d

If using jumbo-frames:

# vmkping -I vmkX <Dest-IP> -s 8972 -d

And should see communication when joining cluster on this:

# tcpdump-uw -i vmkX udp port 12321

"Tried this in chrome and IE11, same thing. Edge, can't even get past step 2 as the next button is gone completely. Firefox, same thing, no next button on step 2."

Have you only attempted this in the UI client or the FLEX/Flash Client also?

It is also relatively trivial to configure a cluster without vCenter aswell if you want to rule out this aspect:

VMware vSphere 6.5 Documentation Library

Bob

View solution in original post

0 Kudos
Rylos
Contributor
Contributor
Jump to solution

@TheBobkin

It was my vCenter version. Was severely out of date. Thought I had updated this when I updated to EP5 on the hosts but had not. Project has been shelved for months and I am just now getting back into it again.

On the old vCenter install step 5 was actually the final step, basically "finish". With updated vCenter step 5 is the witness disks as you said in your reply and step 6 is finish.

I haven't completed the setup yet as I want to update the witness appliance with EP5/6 as well as my hosts but thanks for the direction, it was a huge help.

-Rylos

View solution in original post

0 Kudos
3 Replies
TheBobkin
Champion
Champion
Jump to solution

Hello Rylos

Welcome to Communities and vSAN.

"---2 x 500GB nvme SSD. I didn't realize I couldn't use two separate cache drives unless I have more than one spinning disk. Lesson learned, can't raid, oh well."

Only one cache-tier device per Disk-Group - you could however go with less capacity and better performance and use NVMe devices for both capacity-tier and cache-tier - either way though if this is hardware that is not on the HCL and/or nested then don't be expecting the world with regard to performance and/or potential reliability.

"-Two identical servers running 6.7.0 build 10764712"

"-Witness appliance running 6.7.0 Update 1 (Build 10302608)"

Update your Witness to the same build as your hosts - just because the Witness is an appliance doesn't mean it should be neglected as it is a vital part of the cluster what with being its own Fault Domain.

You never mentioned what build your vCenter is on, if it is 6.7 pre-U1 then upgrade before attempting anything - while 'ESXi' 6.7 U1 + vCenter 6.7 GA are a supported configuration, vSAN 6.7 U1 and vCenter 6.7 GA are not as too much changed and various issues can be encountered even during attempts at initial configuration.

Step 5 as per the 6.7 UI client is claiming the Witness Appliance disks, do you see literally *nothing* or do you see a screen with no disks available for claiming?

You also failed to mention how the Witness has vsan/witness traffic configured, but then again Step 4 in the UI configuration wizard should prompt and not allow selection as cluster Witness if no vmk is enabled for vsan traffic.

"Hosts and witness can talk to each other on all subnets."

The proper way of checking this requires you validate they can communicate over the required interfaces at the configured MTU and on the required port

# vmkping -I vmkX <Dest-IP> -s 1472 -d

If using jumbo-frames:

# vmkping -I vmkX <Dest-IP> -s 8972 -d

And should see communication when joining cluster on this:

# tcpdump-uw -i vmkX udp port 12321

"Tried this in chrome and IE11, same thing. Edge, can't even get past step 2 as the next button is gone completely. Firefox, same thing, no next button on step 2."

Have you only attempted this in the UI client or the FLEX/Flash Client also?

It is also relatively trivial to configure a cluster without vCenter aswell if you want to rule out this aspect:

VMware vSphere 6.5 Documentation Library

Bob

0 Kudos
Rylos
Contributor
Contributor
Jump to solution

@TheBobkin

It was my vCenter version. Was severely out of date. Thought I had updated this when I updated to EP5 on the hosts but had not. Project has been shelved for months and I am just now getting back into it again.

On the old vCenter install step 5 was actually the final step, basically "finish". With updated vCenter step 5 is the witness disks as you said in your reply and step 6 is finish.

I haven't completed the setup yet as I want to update the witness appliance with EP5/6 as well as my hosts but thanks for the direction, it was a huge help.

-Rylos

0 Kudos
Rylos
Contributor
Contributor
Jump to solution

Forgot to add

Have you only attempted this in the UI client or the FLEX/Flash Client also?

I tried from UI and FLEX clients. My guess is that there is something in the older vCenter code that isn't working correctly with the newer browsers or a potential bug.

This is what I was seeing:

vsan1-witness disks.png

vsan2-ready to complete failure.png

0 Kudos