VMware Cloud Community
Jock8186
Enthusiast
Enthusiast

NFS & iSCSI sharing vNic - Is this supported??

I'm at a bit of a logger heads just now with VMware support. My current configuration is as follows:

- Standard vSwitch 1 has 2 vNICs assigned.

- The 2 vNICs have iSCSI offload capability (h/w hba is associated to each vNIC, it's a converged 10GB broadcom card)

- We use both iSCSI & NFS backed datastores

- Within vSwitch 1, vmk2 (nfs, vlan 112) has both vnics configured as active/active

- Within vSwitch 1, vmk3 & 4 are iSCSI Kernels (vlan 102).

- HBA1 is bound to vmk3 (with vnic 1 active and vnic 2 unused)

- HBA2 is bound to vmk4 (with vnic2 active and vnic 1 unused)

The VMware support engineer i have liaising with is stipulating that this is NOT a supported design configuration & that an NFS kernel should not pass traffic through the same vNIC that an iSCSI kernel also utilises (despite both kernels being on a different subnet).

Can anyone clarify this configuration? My understanding is that you should be capable of using various different storage protocols through the vNIC?

Any whitepapers or VMware documentation that can articulate this one way or another would be greatly appreciated.

Many Thanks

Tags (3)
Reply
0 Kudos
2 Replies
ThompsG
Virtuoso
Virtuoso

Hi there Jock8186,

It's a brave man to go against VMware support but when has that ever stopped us Smiley Wink

I've noticed that you have posted the question in both ESXi 5 and 6 forums so not sure what version of ESXi you are actually running. I'm not sure it is important so will continue any way. I'll be upfront and at first glance your configuration looks to be a supported configuration though may not adhere to VMware best practices however that doesn't make it unsupported.

If you take a look at these documents you will see what I mean:

  1. http://www.vmware.com/content/dam/digitalmarketing/vmware/en/pdf/whitepaper/iscsi_design_deploy-whit...
  2. http://www.vmware.com/content/dam/digitalmarketing/vmware/en/pdf/techpaper/vmware-perfbest-practices...

Take a look at the first document on page 8 in the Network Settings section and you see statements like this:

iSCSI best practice.png

As you can see here it talks about best practice and not unsupported. The document goes on to make recommendations about using VLANs for traffic separation and the like.

In second document on page 36 under the section labelled iSCSI and NFS Recommendations you will find statements like this:

iSCSI best practice 2.png

Again no mention of not being supported but again suggests that you use VLANs to separate traffic. In fact this document implies that this is on the same vmknic when it talks about VLAN for the iSCSI/NFS server.

The requirements to make your iSCSI port binding supported are as follows:

  • Array iSCSI ports must reside in the same broadcast domain and IP subnet as the VMkernel port.
  • All VMkernel ports used for iSCSI connectivity must reside in the same broadcast domain and IP subnet.
  • All VMkernel ports used for iSCSI connectivity must reside in the same vSwitch.
  • Port binding does not support network routing.

From what you describe, it appears that you meet the above requirements for iSCSI port binding. You stipulate that iSCSI and NFS are in different subnets however I am assuming with the above that both vmk3 and vmk4 are in the same subnet as your iSCSI storage.

Anyhow, take a look at the documents provided and see if this is helpful. This link also contains some good information either though I'm sure you have already seen it: Considerations for using software iSCSI port binding in ESX/ESXi (2038869) | VMware KB

Kind regards.

ThompsG
Virtuoso
Virtuoso

Hi again,

Sure you also have the covered but another link to look at: Challenges with Multiple VMkernel Ports in the Same Subnet - Support Insider - VMware Blogs This one discusses why you shouldn't have VMkernel ports on the same subnet so probably not relevant but certainly worth a look. Talks about separating the different traffic types like vMotion, Fault Tolerance and IP Storage.


Kind regards.