is it possible to show of serial number of nvme disk by nvmeadm ?

also vir and pid

*vid and pid

The serial number is just in the output of `nvmeadm list`

and in `nvme identify nvmeN`

*nvmeadm

and the PCI IDs are included in `nvmeadm -v identify nvmeN`

andyf: thx, just found serial under real HW. it's show nothing on vmware

another question - can i enable led on under disk socket by nvmeadm ?

we have supermicro box with nvme

my experience has been that vmware's nvme emulation is rather shit tbh

like do NOT try to trim a zpool on it

despite VMware advertising support

it will put your box into a panic loop

(arguably there should be a bug and we shouldn't panic if the dataset mgmt request fails like it does with vmware)

igork: that probably depends on how the LED is controlled.. are the NVMe devices actually showing up as NVMe?

I've seen a number of vendors stick a SCSI<->NVMe adapter (don't know what else to call it) in front of NVMe devices

so they actually show up as SCSI devices

but usually the ident/fault LEDs are managed out of band anyway

so you'd need to know what mechanism is controlling it

(e.g. SES, ....)

jbk: i'm looking real hw server with nvme disks where we can see LED under web ipmi and i'm looking opportunity how to enable LED on device from system

It still depends, unfortunately. For example, we do it through the PCIe slot controls. Some folks did that, others are IPMI, and others are as jbk described too.

with supermicro at least, often their documentation has at least some hints as to what they're using for that

but they're SAS or SATA, not NVMe.

If they do shove a HBA in front of the NVMe devices it might work for you.

jbk: trim works for me without issues on esxi nvme, illumos/freebsd

in the version of esxi here, a virtual NVMe device advertises support but then errors when you actually issue it

[illumos-gate] 15615 SMB2 read should use zero-copy VFS interfaces -- Gordon Ross <gordon.ross⊙tc>

hm, richlowe is somewhere else...

hmm no arekinath... i noticed that the adv_xxx_cap properties for mlxclx always seem to show a default of 0

(I'd think they'd refeclt the default settings of the card)

mlxcx needs some more love for sure, this is one more thing, and it's possible arekinath has fixes for that in his own illumos-joyent branches.

i see there those are getting set, but i'm not familiar enough with it offhand to know what the right answer is

errr where

Me neither. WANTED to dive into mlxcx but can't as quickly as I'd like.

yeah, i've got a number of other things piled up first