ReadyNAS NVX problems with ESXi 5.0 [Case #16549549]

Hey folks. I ran into this on my ESXi 5 box and quickly noticed it was for a SCSI 0x93 command. That's a write_same, which is not quite a command used often for regular VM operation. It's being attempted while either testing VAAI primitives (and repeatedly failing), and/or it's an issue with ESX/ESXi attempting to run array-accelerated commands. I can't really be bothered to drill down into it, but I think it's the former.
You'd run into this usually for your first foray into VM creation, migration, disk creation, etc. on the respective iSCSI LUN, I expect.

Anyway, long story short, disable VAAI (all three accelerated functions!) per the following KB article (http://kb.vmware.com/kb/1033665), try the operation again, and you're good to go.

If you're in a stuck loop with it flooding in your vmkernel logs already, do the nasty and disable iSCSI access via your ReadyNAS for at least five seconds (must result in I/O timeout + retry cycle on ESX), then enable it again.

Edit: I ran into this and got it working normally again, all on the latest release-level firmware, 4.2.19

Edit2: Here's a breakdown for anyone curious.

vmkernel.log from ESXi 5.0:
2011-11-16T03:12:07.518Z cpu3:2051)ScsiDeviceIO: 2305: Cmd(0x4124007ad0c0) 0x93, CmdSN 0x117 to dev "naa.60014052e2150400469d001000000000" failed H:0x0 D:0x2 P:0x0 Valid sense data: 0xb 0x24 0x0.
2011-11-16T03:12:07.523Z cpu3:2051)ScsiDeviceIO: 2305: Cmd(0x4124007ad0c0) 0x93, CmdSN 0x117 to dev "naa.60014052e2150400469d001000000000" failed H:0x0 D:0x2 P:0x0 Valid sense data: 0xb 0x24 0x0.
2011-11-16T03:12:07.528Z cpu3:2051)NMP: nmp_ThrottleLogForDevice:2318: Cmd 0x93 (0x4124007ad0c0) to dev "naa.60014052e2150400469d001000000000" on path "vmhba38:C0:T0:L0" Failed: H:0x0 D:0x2 P:0x0 Valid sense data: 0xb 0x24 0x0.Act:NONE

1) Note the bolded command type, 0x93.
2) Note the Sense Data provided: H:0x0 D:0x2 P:0x0, where D:0x2 means the device/bus was busy (not very helpful yet. But do understand that this will result in ESX retrying this I/O).
3) Note the valid additional sense data/qualifier/key (as opposed to "possible" sense data, which is not necessarily actually received/provided from an array). It says, 0xb 0x24 0x0. You'd need some reference material to figure out what this means. And I don't have it, as I'm not at work.

Looking at Biesser's logs (I don't have SSH enabled on my ReadyNAS yet), I am unsurprised and it all fits together in my head:

beisser wrote:
my system.log shows this: Aug 29 16:13:31 ESX-ISCSI kernel: WRITE_SAME w/o UNMAP bit not supported for Block Discard Emulation Aug 29 16:14:02 ESX-ISCSI last message repeated 8768 times Aug 29 16:14:32 ESX-ISCSI last message repeated 8642 times

- Write same command confirmed.
- unmap / block discard emulation isn't something I have spent time researching, but I presume it's unrelated to VAAI but the two are trying to mingle. So - disable VAAI and we're not going to see ESX/ESXi attempt write_same commands.