VSAN Deployment Questions

[Super6VCA]

We currently have a 5 node ESXi 6 Standard Cluster and in that cluster we have both our server and View VM's mixed in.  This is how it was when i arrived and I now have the chance to change things.  I would like to start utilizing VSAN and had a few questions concerning how to go about this.  First, would it be best to split up the Cluster into 2 separate clusters (1 for View and 1 for servers)?  If it is split, would i be t served to utilize VSAN for the View or Server Cluster?  I'm guessing the view cluster but looking for input.  I am looking to get 3 new hosts that are VSAN Ready but curious if there are any other special considerations that i need to look at when getting this put together.  any comments of suggestions are welcome.  Thanks for reading my post and thanks in advance for any replies. 

Perry

[zdickinson]

First, would it be best to split up the Cluster into 2 separate clusters (1 for View and 1 for servers)?

For licensing purposes, yes.  You need to license each node in a vSAN cluster even if it is not contributing storage.

If it is split, would i be t served to utilize VSAN for the View or Server Cluster?

I would agree, View cluster.  If you only have one shared storage tech for production servers, I would not recommend vSAN.  I like it for VDI, DR, etc...

I am looking to get 3 new hosts that are VSAN Ready but curious if there are any other special considerations that i need to look at when getting this put together.

Stay with unmodified vSAN Ready nodes and you'll be good.  Always make sure your components are on the HCL down to the exact version of firmware and driver.  If possible, add a 4th node so you can rebuild if a host fails.  Look at AFA, with dedupe and compression, it can be cheaper than hybrid.  Take a look at VxRail.

Thank you, Zach.

[Super6VCA]

Zach,

thanks for the reply.  Just curious why you wouldn't recommend vSAN for the production servers?  Can you elaborate a bit??  Thanks again for the response.

[zdickinson]

Good morning, I guess I made some assumptions.  This would be your only production storage and you wear many hats.  Network, storage, VM, backup, etc...  In that case I don't recommend vSAN.  In larger organizations where there are multiple production storage environments and a dedicated storage team that can be assigned to vSAN, yes.

As a do it all admin, I like vSAN in my DR but would not want to use it in production.  However, I am looking at VxRAIL for production which uses vSAN.  I just don't want to manage vSAN in production.

Thank you, Zach.

[elerium]

I've been using VSAN since 6.0 launch and VSAN is production ready and capable. If you have other VSAN environments used for lesser purposes such as Dev/Test or DR, running mission critical loads on VSAN is easy. However I'm also hesitant to recommend VSAN if it's your first and only cluster and it's being used for mission critical workload and you don't have a lab or other environment to test changes. There are bugs (most small but some large and serious) and it takes some time to adjust to  hyperconverged architecture which involves new ways for handling maintenance and upgrades/updates (esp with firmware/drivers/esxi versioning). There are also some under the hood controls that aren't well documented and only learned by opening KB cases with support which in my opinion really should be made user adjustable somewhere in the GUI for troubleshooting and performance tuning /LSOM/lsomLogCongestionHighLimitGB (adjusts write buffering) and /vmkModules/vsan/dom/MaxNumResyncCopyInFlight (adjusts resync) come to mind. These under the hood things become critical when working to replace degraded hardware like failing disks.

In summary, I think VSAN new releases are too buggy to trust with production servers (ESXi 6.0.0 build 4510822 is very solid if you are starting new). The problem comes when you need to patch or update the VSAN version. I've seen major bugs and performance regressions from 6.0 to 6.1 to 6.2 and disk format upgrade to 6.2 was a complete mess for months. To VMware's credit their support has been good and they ultimately have resolved every ticket I've opened but it's very rough running a SAN with performance or stability issues for months with bandaid workarounds while waiting for fixes. I wouldn't attempt any kind of VSAN upgrade today without a test lab or non-critical environment to roll an upgrade to. As an example I upgrade my load test cluster, then dev test, branch offices, prod, lastly DR (waiting weeks between each to see what issues arise). With the bugs I've seen in the last year, upgrading prod VSAN from 6.0 to 6.1 and 6.2 directly would have wrecked havoc in my prod environment.