Microsoft Exchange Server

  • 5 February 2014
  • 8 replies
  • 10251 views

Userlevel 2
Badge +6
Is anyone running Microsoft Exchange on their Nutanix cluster? If so, approximately how many mailbox databases (EDBs) are you running and what is the approximate size of each?

Also, overall feedback/thoughts of the performance of your Exchange environment on the Nutanix cluster?

8 replies

Userlevel 4
Badge +21
I'm using Nutanix with vSphere, so I'm not deploying Exchange, as Exchange is not (yet ?) supported on file storage (even file storage for the hypervisor, like NFS for ESXi, or SMB for Hyper-V), are you using KVM?

If not, for Nutanix/vSphere, see here: http://www.joshodgers.com/2014/02/11/virtualizing-exchange-on-vsphere-with-nfs-backed-storage/
And here for the survey/RFE is here: http://exchange.ideascale.com/a/dtd/support-storing-exchange-data-on-file-shares-nfs-smb/571697-27207

Sylvain.
Userlevel 2
Badge +6
We are using vSphere with NFS storage presented by Nutanix. Understood that there is no official support for Microsoft Exchange backed by NFS, but it was in part ignorance and business decision [ sometimes the two are not different 🙂 ] that we moved forward in this direction. We made the switch to NFS two years ago (early adopters of Nutanix), and our need to be agile and migrate to the Nutanix platform quickly drove us to where we are today.

Luckily, we have not had an issue where we have not been able to get support from Microsoft due to the NFS storage. Overall, our experience has been good. (We have a smallish exchange environment of ~300 - 350 mailboxes)
Userlevel 4
Badge +21
One thing you can try (ask Nutanix first) is to present iSCSI LUNs from Nutanix to host your Exchange VMs on.
As we know Nutanix is capable of presenting iSCSI (for KVM), so I don't think there will be much problem to present some iSCSI LUNs to vSphere.
That way you can migrate just your Exchange VMs to "block" storage and be in line with Microsoft support policy.

Just a thought so, glad to hear that everything is working fine as it is!

Sylvain.
Userlevel 1
Badge +11
Sylvain is correct, you could simply use iSCSI to host Exchange workloads, to be strictly compliant with Microsofts support policy.

However, this would simply add unnessasary complexity to your environment, which is what Nutanix is designed to avoid.

As the various blog / technet posts explain, the issue is not a technical one, so I would recommend customers continue to run Exchange on Nutanix containers presented via NFS.

The support issue is actively being discussed with Microsoft, and we are gaining some traction and hope to pressure MS into updating their support policy.
Badge +5
Hi there rsciaraffo :)

We use Exchange 2007 on Nutanix NX 2000 cluster. We have 2 databases one is 80Gb and the other is 10Gb.

Prior to upgrading to NoS version 3.5.3.1 our experience was mixed. We had daily disconnects between Outlook 2007 and the Exchange server. We expect this on users who have large mail boxes with too many items in one folder (often in the thousands). However it was also happening for users who had small mail boxes with little contents. In addition our Veeam backup software was unable to backup the Exchange VM due to VSS errors.

Now we are on NoS 3.5.3.1 the issues above have gone. In the 2 weeks since we had the upgrade we have not had any users reporting Outlook disconnections. In addition our Exchange vm now backs up successfully every night.

We have seen the articles regarding Exchange on NFS from Microsoft but I would say our recent experience of Exchange on Nutanix is a good one.

Please let me know if I can help any further.
Userlevel 2
Badge +14
We run Exchange 2013 with about 1.6TB of mail split between 30+ databases on a 4-node 6020 cluster using NFS with zero issue. I definitely agree that the Exchange/NFS support statement by Microsoft is more political than technical - there is no need to muck around with iSCSI. The only inconvenient thing about Exchange on Nutanix is that rebooting a CVM causes enough IO Latency to initiate a failover across the Exchange DAG as the hypervisor waits quite a few seconds before directing IO to another Stargate, but it's not like we go around rebooting CVM's every day so it's not that significant of an issue. Its probably a best practice to vacate a node before you initiate a planned CVM reboot anyway. 🙂
Userlevel 2
Badge +15
Tjagoda, please open a support ticket even about the DAG disconnect. I do not consider that acceptable. I am sure Josh Odgers or someone can help fix this on the Exchange side or we can address our HA.py script.

Lets not just settle for acceptable, we want to make our product the best.

Let me know if I can help in any way.

Tony
Userlevel 1
Badge +11
Hi Tjagoda,

I'd like to echo Tony's comments, the issue your experiencing is surprising as this is not what we're seeing at other customers, or personally in my performance lab.

Definatley open a ticket with our support team, and I'll work with them and you to ensure this is resolved.

Reply