VSAN 6.7 and Hypervisor Boot on same controller

[HFMudd]

Mixed mode controllers allow for both RAID and HBA-mode on the same controller. I understand the potential of problems, but I don't see an outright "mixed-mode" being unsupported beyond a couple of considerations in the vSAN 6.7 Administrators Guide and the KB2129050 Best Practices using vSAN and non-vSAN disks with the same storage controller.

My question specifically is with a single controller handling both hypervisor boot (RAID1) and all of other drives in HBA-mode for vSAN, What are the potential issues that would make this configuration a bad idea. Seems modern reputable controllers in this context wouldn't be a problem. I can certainly see stripped sets and/or multiple mixed configurations being a catastrophe waiting to happen.

Thanks,

Jim

[TheBobkin]

Hello Jim,

Welcome to Communities.

"What are the potential issues that would make this configuration a bad idea"

As the kb article clearly outlines in firetruck red:

"Until IO controllers are formally certified as supporting this dual mode, you should be aware that there is a risk that errors occurring in one domain ( For Example: vSAN) can affect, and hang, the other domain (For Example: VMFS) causing the node to go offline."

and

"Mixing the controller mode will mean that various disks will be handled in different ways by the storage controller. This introduces the possibility that issues affecting one configuration could also affect the other, with possible negative consequences for vSAN."

"Seems modern reputable controllers in this context wouldn't be a problem."

Reputable or not, these and their drivers are generally designed around predictable workload of defined types and adding extra complexity that these may not be specifically designed for is probably not a great idea.

Also, this support statement (yes I know it says RAID0 not RAID1 but still):

https://kb.vmware.com/s/article/53573

Do you have some particular issue with using SD-cards/SATADOM/M.2-BOSS(if Dell)? Will save you two slots this way too.

Bob

[HFMudd]

Hi Bob, thank you for the reply.

To be honest the quotes you have I have read, but there is a lot of ambiguity and that is where I find myself. There is a "risk" but that's kind of a bucket like hyper-converged is a "risk" is from a certain perspective. I keep hearing mixed-mode is "unsupported", but I don't see that, I see there is "potential risk", I see you "you should consider" and other grey terms on the subject in the documentation. Certainly HCL certification would be meaningful. I would like to know where how that will be listed in the HCL. I went through several controllers from different vendors and I didn't find any so either there are no certified controllers or I just don't know how to find that particular certification attribute in the HCL.

As far as booting from USB, we do that with other hosts but in our mission critical environment the ready options available to us at the moment to have RAID1 for the OS is via small SSD pair. If this customer's configuration is truly "unsupported" then I'll do what I need to do to ensure their supportability.

I hope that makes sense.

[elerium]

This mostly comes from issues with the Dell H730 controller. I have seen firsthand the behavior of mixed controller (RAID for boot, passthrough for VSAN) where a power on disk reset for the RAID volume ends up impacting and downing the VSAN passthrough portion, effectively dropping the disk group for a host.

The same best practices KB used to say that mixed was supported except when using H730. My guess is that similar problems have surfaced with other controllers and now they need to qualify individual controllers for mixed mode. Am going to guess that similar issues will occur for mixed mode on any controller using LSI 3108 (chipset used on H730). This KB has a list of other controllers using the same chipset Required vSAN and ESXi configuration for controllers based on the LSI 3108 chipset (2144936)​.

[TheBobkin · Accepted solution]

Hello HFMudd​,

I am presently not aware of any timeline regarding controller certification specifics (and likely couldn't share this) but did mean to get more info.

Any ambiguity you determine is based on the fact that depending on what is used (and how) determines how stable and supportable any system can be - as you can imagine with all the possible combinations, a blanket statement of 'Yes' would belie reality e.g. someone trashing unsupported consumer-grade disks with workload and then questioning why it is non-functional, there has to be lines in what is feasible to support. Me, I do always try to help those in 'unsupported' configurations but there is a line of what will work and what doesn't, unfortunately this is line is grey as there are a lot of variables here - adequately test whatever you are planning on running if you consider it critical workloads.

Whether using what you suggested or something else, if possible do work with the hardware vendor(s) and PSO to validate any configuration that isn't clear regarding support (and from whom).

Bob

[HFMudd]

Thanks Bob and Elerium for entertaining the question.