Thank you for confirming it was indeed still working.
I spend most of my evening tracking down the problem.
The problem was that I had cpu_type set to "kvm64,+aes" when I changed
it a while ago from "qemu64" to expose the AES feature (as the box does
lots of TLS/SSL) I switched to kvm64.
That broke the acpi shutdown. If you look at "qemu-system-x86_64 -cpu
?dump" there is only a very small difference. popcnt sse4a abm svm
lahf_lm are missing with kvm64 but present in qemu64.
Switching to "qemu64,+aes" made OpenBSD respond to the acpi shutdown
command again!
I did some more tinkering and finally settled on "qemu64,+x2apic,+aes",
the x2apic seems to lower qemu's CPU usage from a constant 4% or up to
around 2% or up!
Regarding to qemu_opts property in vmadm... you might want to add "-vga
none" (I have "-parallel none -k en-us -vga none -nographic").
vmadm won't allow you to set the vga to anything other than
std,vmware,qxl and circus. But the last -vga encouters wins. This does
not only disable the vnc console it actually removed the pci bus device.
There is now no mention at all of wsdisplay in dmesg.
Also posting to the list again, as this might be useful for someone down
the road.
Regards
Jorge
On 2016-06-07 19:33, Tamas Gerczei wrote:
Curious - it still works on my 5.9, I've just tested. Admittedly I'm
overriding the default value of "qemu_opts" with "-parallel none -k
en-us -nographic" though.
Jorge Schrauwen wrote:
I already have that enabled as it use to work fine (around 5.6ish I
think).
$ rcctl get apmd
apmd_class=daemon
apmd_flags=
apmd_timeout=30
apmd_user=root
But currently on 5.9 it is ignore the acpi shutdown command.
Not the biggest of deals though as I usually just stop the vm before
updating my SmartOS pi.
On 2016-06-06 22:29, Tamas Gerczei wrote:
Slightly off-topic, but a direct response regarding ACPI vs OpenBSD:
this is only true if you don't run apmd. (Which is supplied, so
adding
'apmd_flags=""' to '/etc/rc.conf.local' and then starting it via
'/etc/rc.d/apmd start' will do.)
Jorge Schrauwen wrote:
I generally use 'shutdown -g0 -i5 -y' for poweroff en 'shutdown -f0
-i6
-y' for reboots and it stops the zones.
KVM zones get a ACPI shutdown signel, from my experience windows
does
the right thing, linux too if acpi stuff is installed. OpenBSD just
ignores it unfortunately.
---
~ sjorge
On 2016-06-06 19:53, Paul Sture wrote:
On 6 Jun 2016, at 19:40, Gareth Howell wrote:
I’m monitoring this with interest.
If I issue a shutdown to the global zone on the server, will the
zones have had the chance to do a graceful shutdown before the
global zone shuts down?
I do regular shutdowns from the global zone and in my experience
'poweroff' waits until all the zones have done a graceful shutdown.
In the last year I've had a couple of instances where a zone hasn't
closed in response to a 'poweroff' and either hung or provoked a
timeout error. Sorry I didn't research the reasons for the timeout
- I
managed to ssh into the offending zone and do a poweroff inside
that,
then reissued the GZ poweroff.
-------------------------------------------
smartos-discuss
Archives: https://www.listbox.com/member/archive/184463/=now
RSS Feed: https://www.listbox.com/member/archive/rss/184463/25769125-55cfbc00
Modify Your Subscription:
https://www.listbox.com/member/?member_id=25769125&id_secret=25769125-7688e9fb
Powered by Listbox: http://www.listbox.com
