VSAN 6.5 causes magic HBA queuedepth upgrade? (LSI SAS 9207-8i)

[srodenburg]

Hello,

We run LSI SAS 9207-8i HBA's (with FW19 as FW20 is a desaster) and they have a queue-depth of 895. Well... at least up until VSAN in ESXi 6.0 U2 that is, because after the upgrade to ESXi 6.5.0a they suddenly have a queue-depth of 8320 on all VSAN node hosts :smileyshocked:

Say wut ?

Yes it's true. In esxtop, type "d" to go to Adapters, type "f" to add a field. Type "d" to add the "queuestats" field and go back. Instead of "AQLEN" being 895 (as it has aaaalways been) it now shows 8320.

As Trump would say "it's huuuuuge" :smileysilly:

So ehm, "whoohoo thank you VMware" or "wft is going on here"?

Can anyone shed some light on this?

[Jasemccarty]

srodenburg​

Not sure why it is showing the high queue depth.

Also, I don't believe the 9207-8i is on the vSAN 6.5 VCG. It only appears to be certified through 6.0 Update 3.

Dell: VMware Compatibility Guide - I/O Device Search  - Dell 9207-8i

LSI: VMware Compatibility Guide - I/O Device Search - LSI 9207-8i

The vSAN VCG can be found here: http://vmwa.re/vsanhclc

[GreatWhiteTec]

Jase is correct. That controller is not certified for vSAN 6.5, neither is FW20.

[srodenburg]

Yes I know that this HBA is not on the 6.5 HCL for VSAN (yet). I started two threads here about exactly that topic:

First one:

LSI Controllers and 6.5 HCL  (only 1 card is listed, all others are gone)

As stated, the 6.5 list of supported is very short at the moment. LSI for example only have one card in the list.

This is the other topic:

LSI SAS 9207-8i and vSAN 6.5 ?​

VMware (or the card manufacturers) cannot leave it like this, because there are many installations out there and all those partners and all those customers cannot suddenly start yanking out cards out of all of their VSAN nodes, that used to be fully supported and whooosh after the upgrade not any longer. That is bollocks.

Sorry guys, with all due respect, but your replies are the kind of useless answers I expect from unexperienced support-personell.

Maybe I stubled upon a bug. I cannot imagine FW19 being the cause and needing FW20. FW20 was a piece of unstable shit in VSAN 6.1 and VSAN 6.2 and was not certified as a consequence (only FW19 was certified). I had to downgrade 5 cards from 20 to 19 because nowadays, cards being shipped have 20 on them.

[Jasemccarty]

Your perspective is appreciated @srodenburg.

Keep in mind that the LSI 9207-8i is similar to the HP P220, but they are different cards. I'm not specifically familiar with the firmware on these cards, but I will state that HP firmware could be different than the LSI firmware.

I have some Dell H730 Mini cards (LSI 3108 based) and some LSI 9361-8i cards (LSI 3108 based), and their firmware is indeed different.

While the H730 cards support Pass through for both Hybrid and All-Flash configurations, the 9361-8i's only support pass through for Hybrid configurations (RAID0 for All-Flash). The H730's are approved for 6.5, while the 9361-8i's are not. Also, the H730's are more common than 9361-8i's.

My point, is that each card has to be individually validated. There are a large number of cards to test and validate on different builds of vSAN. There are a lot of factors that go into which cards are certified and in what order they are in the process.

As far as the queue depth indicating an unexpected value goes, it could possibly be a bug. Specific to that, the recommendation would be to file a ticket.