#SAN vs NAS limits in HWU

1 messages · Page 1 of 1 (latest)

fossil chasm
#

When HWU shows differing limits for SAN and NAS for a particular configuration item (like Flexvol count -- in older versions of ONTAP -- or maximum number of nodes), what is the setting within ONTAP that makes ONTAP say "This is a SAN cluster now" and apply SAN limits versus the NAS limits? (This assumes ONTAP starts as NAS and "changes" to SAN rather than the other way around. But knowing the history of ONTAP, this seems like a safe assumption.)

Installing a block protocol license? Configuring an SVM with a block protocol? Configuring a LIF with a block protocol? Provisioning a LUN?

I'm pretty sure it's not the first two. The second two seem too late. Is there something else I'm missing?

tiny walrus
#

(Most of) these limits are "soft" limits that are not code-enforced. You can have a 24 node cluster that serves LUNs, for example. Or have more volumes than what the HWU suggests.
This only means that you're crossing into "danger" territory. Especially for SAN this can mean: more paths to your LUNs than what a host can handle (I have seen hosts crash if they see more than 16 paths to a LUN, for example). Difficulty of keeping the takeover/giveback times below the timeout limit for SAN, etc

#

usually you should request a PVR for such use-cases, otherwise it could happen that NetApp support can't/won't help you if they are convinced that an issue you're having stems from exceeding these limits

#

I know NetApp had a PDF somewhere about those limits, whether they were per-node, per-HA-pair or per-cluster, and whether each limit was code-enforced or not. I don't know if that still exists?

fossil chasm
#

My question is more around monitoring and detection, not necessarily that I intend to exceed any of these limits. I want to alert before those thresholds are crossed but first I need to detect the relevant parameters so I can apply the proper limits. Vol count is a hard limit and it's a limit I've reached before; and with the version of ONTAP I have some places the difference still matters. (Newer ONTAP versions are bringing increasing parity between the cluster types so the difference matters less.)

pliant osprey
#

Also if we have to go to Engineering for a Support case and find a config not meeting HWU/IMT specs, Engineering will refuse to work on it without a PVR, or likely will.

tiny walrus
# pliant osprey Also if we have to go to Engineering for a Support case and find a config not me...

...true, but only if they suspect that the breached limit is in any way related to the issue you're having. It's not like they're checking all limits before working on your case 🙂 We had Support (1st and 2nd level) happily help us troubleshoot and fix an issue in a feature which was not even publicly documented outside of an EAP program, yet the customer was using it on his production system 🙂

pliant osprey
#

Eng != L2. But if we say no don't be too surprised.

#

We try to be customer friendly but we do have to enforce some thing unfortuantely.

tight belfry
# tiny walrus ...true, but only if they suspect that the breached limit is in any way related ...

It isn't even if they suspect it. Speaking from experience on this. We had a cluster with more than the 'supported' number of disks and ran into a problem with random disconnects from a SAN host.
started working with support and shortly after we pulled up the screen showing the system status she put me on hold and came back with a manager saying they couldn't do anything with the system being 'out of support spec' and as soon as that was corrected to open a new case.
thankfully, our account team already had an engineer working on our system for a POC and after sorting that out we got everything fixed (it was windows, not netapp at fault.)
but yea, they can shut you down pretty quick. Not 'will' but can.

not sure on the alerting thing, unless you use something such as NABox, AIQ or a custom python/rest/PS/etc script that runs on a schedule.
From the system itself, I don't know of anything that will alert as it isn't really coded into ONTAP

torpid echo
#

I find that odd. Last time I tried, I had a fas 8200 that would not boot since it had more than the supported drive count. It’s coded into ONTAP

tight belfry
#

we were working with CVO

tiny walrus
tiny walrus