Hello All, im just updating a legacy thread on Proxmox KVM Hosts for OpenBSD Guuests The KVM / Proxmox 5.x Preemption timer in Linux Kernel In proxmox 5.3 ... OpenBSD 6.4 Guests work fine without any modifications to the Proxmox System, (ie you no lonter need to disable the) /sys/module/kvm_intel/parameters/preemption_timer
so OpenBSD Runs fine on an unmodified Proxmox 5.3 Box ... so the KVM / Linux Bug that caused OpenBSD guests to Freeze seems to be resolved in Proxmox 5.3 ... (so users of 5.0 , 5.1 may want to upgrade to the latest version of Proxmox I hope this helps Tom Smyth On Mon, 15 Jan 2018 at 21:34, Tom Smyth <tom.sm...@wirelessconnect.eu> wrote: > Hello all, I just want to reference the following post which resolved > my issues of running > OpenBSD guests on Proxmox 5.x > > Just to clarify Todds / Stefans Email in another thread (Thanks > Stefan and Todd,) > > https://marc.info/?l=openbsd-misc&m=151600765414976&w=2 > > > disable kvm_intel.preemption_timer on the host (see \ > > /sys/module/kvm_intel/parameters/preemption_timer ) This seems to be > buggy in linux \ > > 4.10 and newer > > disabling intel-kvm preemption_timer worked for me and it resolved the > timer drift issue > in openbsd on Proxmox 5.1 > > so in debian / proxmox I ran the following command > > echo options kvm-intel preemption_timer=N >>/etc/modprobe.d/kvm-intel.conf > Thanks, > > Tom Smyth > > On 30 December 2017 at 23:25, Tom Smyth <tom.sm...@wirelessconnect.eu> > wrote: > > Hello I have repeated OpenBSD 6.2 Testng on Proxmox PVE > > 5.1 3r Relasee CD > > > > The console does not hang like in previous releases of Proxmox > > PVE 5.x > > However the issue of delays between pings (slow sleep time) stil > > is there since the 5.1 Release 2 and is present in 5.1 release 3 iso > > (the Proxmox 5.1 CD that was released 22 December 2017 > > > > if I do date;sleep 1;date > > I will get the first time and date, and the second time about > > 9-11 seconds after the first... and the interval between pings is > > sporadic... I will raise a case with Proxmox again about this > > Ill do some further digging... > > Thanks > > > > On 27 October 2017 at 07:18, Tom Smyth <tom.sm...@wirelessconnect.eu> > wrote: > >> Hello Theo, Mike, All, > >> > >> @Theo Understood it is important to protect developers and the project > goals > >> ... @Mike Thanks for your Generosity in the time you took on this > thread, > >> Yes I want Mike to make VMM more awesome :) @Mike keep up the good work > >> > >> I cant disagree with any point that Theo made in his email on this tread > >> that said, > >> unfortunately I cant always choose my hypervisor and I dearly want to > run > >> OpenBSD on it proxmox... > >> > >> I do think (based on the fact that OpenBSD 6.0-6.2 works on PVE 4.4 it > is > >> probably a (virtual Hardware issue ) .. not necessarily an OpenBSD issue > >> I will raise this with the PVE Support guys (as I have already done > since mid > >> July ) > >> > >> Any further posts on this thread from me will be (hopefully for other > OpenBSD > >> users benefit (if I make progress) > >> and certainly not intended as a request or a distraction for Core > >> OpenBSD Developers > >> > >> All the Best, > >> > >> Tom Smyth > >> > >> On 27 October 2017 at 06:37, Theo de Raadt <dera...@openbsd.org> wrote: > >>> Tom, > >>> > >>> A virtual machine setup is an operating system running on an operating > >>> system on top of an operating system. > >>> > >>> OK, not quite. The middle one, the VM itself, is as a bit less > >>> complex than a full operating system as machine-independent code goes, > >>> but nevertheless the machine-dependent bat-shit-crazy stuff is far > >>> more complex with gobs of extremely messy nuances face it on both > >>> sides because x86 is a fucking minefield > >>> > >>> Everyone needs to adjust their expectation that all 3 layers are > >>> perfect, AND not assume that it is our layer doing the wrong thing > >>> > >>> Really the layers should simplify but the current marketplace is still > >>> gaining more value out of product differentiation than > >>> simplification+convergence, both sw and hw > >>> > >>> Even if our subsystem isn't doing something 'right', it is NOT the > >>> stated goal of OpenBSD to run well on every garbage VM, because it has > >>> become impossible for the little guy to be perfect. > >>> > >>> Concerted efforts to diagnose and improve these low-level issues uses > >>> the same crowd of people who are trying to improve other edges which > >>> may be more important. do you want our vmm to work well? or do you > >>> want us to work better on someone else's vmm? Sorry, limited > >>> skillset, pick what you want mlarkin to focus on! But that is unfair, > >>> and even if he listened to your wishlist, UNPRODUCTIVE. > >>> > >>> Where does this go? Get ready for monopolies in everything, or > >>> oligopolies at best... or fight their establishment. > >>> > >>>> Just to say the gaps in ping response seems get worse as the uptime > increases > >>>> ie > >>>> with the uptime around 5 minutes the gaps between ping results are > around 1 sec > >>>> (what I consider normal) > >>>> with the uptime around 2 hrs 45 minutes the gaps between ping results > are 13 sec > >>>> with the uptime 8 hrs 30 minutes the gaps between ping results are > 35 seconds > >>>> > >>>> Output of sysctl kern.timecounter below > >>>> > >>>> kern.timecounter.tick=1 > >>>> kern.timecounter.timestepwarnings=0 > >>>> kern.timecounter.hardware=acpihpet0 > >>>> kern.timecounter.choice=i8254(0) acpihpet0(1000) acpitimer0(1000) > >>>> dummy(-1000000) > >>>> > >>>> I will change the ACPI now to i8254 and report back later on > >>>> Thanks > >>>> > >>>> > >>>> On 26 October 2017 at 20:25, Mike Belopuhov <m...@belopuhov.com> > wrote: > >>>> > On Thu, Oct 26, 2017 at 19:05 +0100, Tom Smyth wrote: > >>>> >> Lads, > >>>> >> > >>>> >> Im pleased to say that my testing of OpenBSD 6.1 and OpenBSD 6.2 > >>>> >> Release > >>>> >> amd64 , > >>>> >> appear to work a little better in Proxmox PVE5.1 as released > this week, > >>>> >> > >>>> >> I used iso version 5.1-722cc488-1 from Proxmox > >>>> >> Updated on 24 October 2017 > >>>> >> > >>>> >> The Console no longer freezes but after a few hours > >>>> >> the console (vga console accessed via Proxmox webinterface seems > >>>> >> to lag a little > >>>> >> the interval between pings for instance takes up to 13 seconds, > which > >>>> >> is a bit strange... ie it takes 13 seconds for each line of Ping > result > >>>> >> which is u > >>>> >> Ill report more feedback later, but at least OpenBSD is not > freezing > >>>> >> as bad in this > >>>> >> version of Proxmox PVE 5.1 > >>>> >> > >>>> > > >>>> > Hi, > >>>> > > >>>> > Can you please show us the output of "sysctl kern.timecounter". > >>>> > If you're currently using an acpihpet0, can you please try > >>>> > switching to the acpitimer0 (and if that doesn't help, i8254) via > >>>> > > >>>> > sysctl kern.timecounter.hardware=acpitimer0 > >>>> > > >>>> > and attempt to reproduce the 13 secod delay. > >>>> > > >>>> > Regards, > >>>> > Mike > >>>> > >>> > >
