-
lundman
not seeing disks on new hardware: pci15d9,1b4b
-
lundman
any hope?
-
kebe
lundman: pci15d9,1b4b is the motherboard one.
-
kebe
try this: `prtconf -d | grep 1000` or `prtconf -vp | grep 1000` I'm assuming you've an HBA of some sort on this supermicro (pci 15d9) machine?
-
» kebe is hoping 1000 matches because pci1000 is LSI-now-Broadcom.
-
lundman
i took a guess, in data centre.. looking now
-
lundman
grep for not attached might work too
-
lundman
ok 15d9,1b99
-
lundman
pciex 1000,e6
-
kebe_
Aha!
-
kebe_
THAT is someting I can work with.
-
lundman
you called it
-
lundman
sas38xx
-
kebe_
-
kebe_
Okay, first test...
-
kebe_
`grep 1000,e9` /etc/driver_aliases
-
kebe_
ooops e6 I meant
-
kebe_
Likely no entry.
-
lundman
none
-
kebe_
This is OmniOS and you've a root pool that's NOT attached to the HBA, right?
-
kebe_
Or is this the installer?
-
lundman
installet 44
-
kebe_
Hmmmm...
-
lundman
highest 1000 is d2
-
kebe_
can you *edit* and save-out /etc/driver_aliases? (Is it like ramdisk root? Been a while since I've played with the OmniOS installer.)
-
lundman
yes
-
kebe
Okay, add a line like one of the other mpt_sas entries, but for 1000,e6.
-
kebe
Make sure quotes are in the right places, etc.
-
lundman
yep five
-
lundman
err five
-
lundman
done
-
lundman
stood phone
-
kebe
huh? "stood phone"?
-
lundman
sorry my phone auto corrected done to five
-
lundman
running disks
-
kebe
Okay.
-
lundman
command
-
kebe
No.
-
kebe
not yet.
-
lundman
ok
-
kebe
First you utter "devfsadm" which will re-read the driver_aliases file. THIS MIGHT NOT WORK. THIS MIGHT EVEN PANIC YOUR KERNEL.
-
kebe
You're the first person I know with a 38xx chipset to try this on.
-
kebe
Good luck.
-
lundman
ran
-
kebe
Got your shell back?
-
kebe
Any console noise even?
-
lundman
yeah complaining about other things
-
kebe
Hmmmm... like what?
-
lundman
removing link. pcie0
-
kebe
Huh...
-
lundman
347c invalid contents
-
kebe
But you have your shell?
-
kebe
Is that from mpt_sas or some other driver?
-
lundman
shell ok
-
lundman
i think unrelated
-
kebe
Try diskinfo now?
-
lundman
only usb
-
kebe
Shit. What does "modinfo | grep mpt_sas" have to say?
-
lundman
wish i took bloody version
-
kebe
Probably wouldn't have helped you.
-
lundman
mptsas loaded 191 00.00.00.24
-
kebe
SHould look more like:
-
kebe
249 fffffffff83d9000 37bd0 185 1 igb (Intel 1Gb Ethernet)
-
kebe
but "mpt_sas" and differenet values.
-
lundman
yes it does
-
kebe
Okay. So the driver loaded...
-
kebe
Wait.
-
lundman
prtconf has it attached as well
-
lundman
just no disks
-
kebe
`cfgadm -l ` (might need to pipe it into a pager).
-
kebe
Anything pop up there for scsi/SAS ?
-
lundman
no cfgadm in installer
-
kebe
Oh damn.
-
kebe
Okay... if you can find someone to ROLL you an installer that has THAT line added in /etc/driver_aliases you can try again.
-
kebe
I'd normally offer to roll a SmartOS USB or ISO but I've my own issues with getting a release out this week and helping RM with the IPD 38 YMM work.
-
lundman
thanks. my time is up here, will ask them to send a card for us to play with
-
lundman
thank you for your help
-
kebe
You're not the only one asking about the 9500 with the 38xx chipset in it. nomad has skin in that game too.
-
kebe
Part of me thinks just adding the line to driver_aliases will be sufficient, from what I recall seeing in the freebsd equivalent of mpt_sas, but it's been long enough where I'm not 100% sure.
-
lundman
yeah it certainly seemed happy to attach it
-
kebe
My SAS-fu is weak. We use NVME or SATA in-house at MNX, and I use SATA at home on both HDC (OmniOS) and Kebecloud (Triton/SmartOS).
-
kebe
It's past midnight here, I'm off to bed. (And tomorrow I have to run more tests on YMM before getting ready to cut this week's SmartOS.)
-
nomad
kebe, we went ahead and ordered the new file server with both a 3008 and a 9500 card. Hopefully we'll be able to use the 9500 soonish.
-
nomad
I'm going to try to find a way to get you access to a host with the 9500 (and the 3108 :) installed for when you (or someone else with devskill) has time to look at it.
-
richlowe
when does bloody update its illumos?
-
hadfl
usually on thursday
-
pmooney
is there a plan to push updated versions of the onbld package with the gcc10-as-primary env files?
-
pmooney
I saw that had been backported to the various branches
-
hadfl
yes the packages should land after tomorrow's merge. let me know if you need them more urgently
-
hadfl
that is for bloody. for the releases they'll be published with the next release which has no due date, yet
-
pmooney
no urgency on bloody
-
pmooney
considering gcc10 builds are expected for upstreaming now, does it make sense to get an updated package out for the releases?
-
richlowe
I had no idea anyone at all used the onbld env files.
-
richlowe
rather than the ones in usr/src/tools/env
-
richlowe
that's certainly good to know, because I need to sort out the onbld mess one day
-
hadfl
we can certainly get a release out for just onbld for all releases rather than waiting for more things before cutting a new release
-
hadfl
downloads.omnios.org/env << there are also up to date .env files to grab from here
-
pmooney
richlowe: I use the omnios ones since they have the necessary local overrides
-
pmooney
hadfl: I don't know what all is involved with cutting a release (more than pushing an updated package for developer/onbld), but considering that using those env files is the documented way for doing builds per the omnios site, it'd be nice for them to be in a configuration "acceptable" for the gate
-
pmooney
I have no problem modifying my own local configs. I'm just thinking about anyone else who walks up to the project