11:47:12 <igork> is it possible to show of serial number of nvme disk by nvmeadm ?
11:47:24 <igork> also vir and pid
11:47:35 <igork> *vid and pid
11:53:16 <andyf> The serial number is just in the output of `nvmeadm list`
11:53:38 <andyf> and in `nvme identify nvmeN`
11:54:16 <andyf> *nvmeadm
11:54:31 <andyf> and the PCI IDs are included in `nvmeadm -v identify nvmeN`
13:12:01 <igork> andyf: thx, just found serial under real HW. it's show nothing on vmware
13:12:56 <igork> another question - can i enable led on under disk socket by nvmeadm ?
13:13:08 <igork> we have supermicro box with nvme
13:19:05 <jbk> my experience has been that vmware's nvme emulation is rather shit tbh
13:19:26 <jbk> like do NOT try to trim a zpool on it
13:19:32 <jbk> despite VMware advertising support
13:19:41 <jbk> it will put your box into a panic loop
13:20:01 <jbk> (arguably there should be a bug and we shouldn't panic if the dataset mgmt request fails like it does with vmware)
13:21:40 <jbk> igork: that probably depends on how the LED is controlled.. are the NVMe devices actually showing up as NVMe?
13:22:13 <jbk> 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
13:22:21 <jbk> so they actually show up as SCSI devices
13:22:43 <jbk> but usually the ident/fault LEDs are managed out of band anyway
13:22:52 <jbk> so you'd need to know what mechanism is controlling it
13:22:57 <jbk> (e.g. SES, ....)
13:24:13 <igork> 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
13:40:02 <rmustacc> 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.
14:25:04 <jbk> with supermicro at least, often their documentation has at least some hints as to what they're using for that
14:58:49 * nomad uses sas3ircu to control drive lights on his SuperMicro fileservers.
14:59:02 <nomad> but they're SAS or SATA, not NVMe.
14:59:46 <nomad> If they do shove a HBA in front of the NVMe devices it might work for you. 
15:17:31 <yuripv> jbk: trim works for me without issues on esxi nvme, illumos/freebsd
16:01:02 <jbk> in the version of esxi here, a virtual NVMe device advertises support but then errors when you actually issue it
16:40:18 <gitomat> [illumos-gate] 15615 SMB2 read should use zero-copy VFS interfaces -- Gordon Ross <gordon.ross⊙tc>
19:15:34 <tsoome> hm, richlowe is somewhere else... 
19:26:52 <jbk> hmm no arekinath... i noticed that the adv_xxx_cap properties for mlxclx always seem to show a default of 0
19:27:15 <jbk> (I'd think they'd refeclt the default settings of the card)
20:18:36 <danmcd> 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.
20:20:58 <jbk> i see there those are getting set, but i'm not familiar enough with it offhand to know what the right answer is
20:21:04 <jbk> errr where
20:26:03 <danmcd> Me neither.  WANTED to dive into mlxcx but can't as quickly as I'd like.
20:49:32 <jbk> yeah, i've got a number of other things piled up first