Aleix Dorca Josa
2012-Aug-27 06:29 UTC
Clock problems on Squeeze/Xen4 (clock jumps 50 minutes ahead)
Hi all, My xen farm is plagued with this ''Jump 50 minutes ahead'' problem. All my machines run Debian Squeeze and Xen 4 in Dom0. DomU''s are Squeeze too. I thought this was a hardware problem but then I saw this (kind of old) messages: http://lists.xen.org/archives/html/xen-users/2011-02/msg00609.html http://bugs.debian.org/cgi-bin/bugreport.cgi?bug=599161 After reading all available info I still don''t know if the bug has been resolved in later builds of either the kernel or xen on Squeeze. Has anyone got new information on this? I see that the clocksource=pit on the grub command line does not seem to fix the problem. Any other hints? Change to Ubuntu server? Build my own xen binary? Use Debian''s backports to change the kernel version? This is kind of a big problem for me since the error kills ntpd on DomU''s and then Dovecot''s processes stop themselves when time is corrected. Scripts to workaround this are fine but I would rather much prefer for the clock not to jump ahead. Thanks a lot, Aleix.
Stephan Seitz
2012-Aug-27 08:30 UTC
Re: Clock problems on Squeeze/Xen4 (clock jumps 50 minutes ahead)
Am Montag, den 27.08.2012, 08:29 +0200 schrieb Aleix Dorca Josa:> I see that the clocksource=pit on the grub command line does not seem to fix the problem. Any other hints? Change to Ubuntu server? Build my own xen binary? Use Debian''s backports to change the kernel version? > > This is kind of a big problem for me since the error kills ntpd on DomU''s and then Dovecot''s processes stop themselves when time is corrected. Scripts to workaround this are fine but I would rather much prefer for the clock not to jump ahead.Hi, as a quick fix, you could handle your domU''s independent from xen: echo "jiffies"> /sys/devices/system/clocksource/clocksource0/current_clocksourceecho "1" > /proc/sys/xen/independent_wallclock take a look at /sys/devices/system/clocksource/clocksource0/available_clocksource to see which clock sources are possible. cheers, Stephan _______________________________________________ Xen-users mailing list Xen-users@lists.xen.org http://lists.xen.org/xen-users
Aleix Dorca Josa
2012-Aug-27 08:38 UTC
Re: Clock problems on Squeeze/Xen4 (clock jumps 50 minutes ahead)
Hi Stephan, thanks for your suggestion but unfortunately what you propose only works in Xen 3. I am using Xen 4. cheers, Aleix. El 27/08/2012, a les 10:30, Stephan Seitz <s.seitz@netzhaut.de> va escriure:> > > Am Montag, den 27.08.2012, 08:29 +0200 schrieb Aleix Dorca Josa: > >> I see that the clocksource=pit on the grub command line does not seem to fix the problem. Any other hints? Change to Ubuntu server? Build my own xen binary? Use Debian''s backports to change the kernel version? >> >> This is kind of a big problem for me since the error kills ntpd on DomU''s and then Dovecot''s processes stop themselves when time is corrected. Scripts to workaround this are fine but I would rather much prefer for the clock not to jump ahead. >> > > Hi, > > as a quick fix, you could handle your domU''s independent from xen: > > echo "jiffies" > /sys/devices/system/clocksource/clocksource0/current_clocksource > echo "1" > /proc/sys/xen/independent_wallclock > > take a look at /sys/devices/system/clocksource/clocksource0/available_clocksource to see which clock sources are possible. > > cheers, > > Stephan
Steven Timm
2012-Aug-27 14:05 UTC
Re: Clock problems on Squeeze/Xen4 (clock jumps 50 minutes ahead)
Are you running 32-bit domU''s or 64-bit? I have seen the problem many times under xen 3.1.2 but only with 32-bit guests on 64-bit hosts. Steve timm On Mon, 27 Aug 2012, Aleix Dorca Josa wrote:> Hi all, > > My xen farm is plagued with this ''Jump 50 minutes ahead'' problem. All my machines run Debian Squeeze and Xen 4 in Dom0. DomU''s are Squeeze too. I thought this was a hardware problem but then I saw this (kind of old) messages: > > http://lists.xen.org/archives/html/xen-users/2011-02/msg00609.html > http://bugs.debian.org/cgi-bin/bugreport.cgi?bug=599161 > > After reading all available info I still don''t know if the bug has been resolved in later builds of either the kernel or xen on Squeeze. Has anyone got new information on this? > > I see that the clocksource=pit on the grub command line does not seem to fix the problem. Any other hints? Change to Ubuntu server? Build my own xen binary? Use Debian''s backports to change the kernel version? > > This is kind of a big problem for me since the error kills ntpd on DomU''s and then Dovecot''s processes stop themselves when time is corrected. Scripts to workaround this are fine but I would rather much prefer for the clock not to jump ahead. > > Thanks a lot, > > Aleix. > _______________________________________________ > Xen-users mailing list > Xen-users@lists.xen.org > http://lists.xen.org/xen-users >------------------------------------------------------------------ Steven C. Timm, Ph.D (630) 840-8525 timm@fnal.gov http://home.fnal.gov/~timm/ Fermilab Computing Division, Scientific Computing Facilities, Grid Facilities Department, FermiGrid Services Group, Group Leader. Lead of FermiCloud project.
Aleix Dorca Josa
2012-Aug-27 14:17 UTC
Re: Clock problems on Squeeze/Xen4 (clock jumps 50 minutes ahead)
Nope, all is 64 bits (Dom0 and DomU''s). Aleix. El 27/08/2012, a las 16:05, Steven Timm escribió:> Are you running 32-bit domU''s or 64-bit? I have seen the problem many times under xen 3.1.2 but only with 32-bit guests on 64-bit hosts. > > Steve timm > > > > On Mon, 27 Aug 2012, Aleix Dorca Josa wrote: > >> Hi all, >> >> My xen farm is plagued with this ''Jump 50 minutes ahead'' problem. All my machines run Debian Squeeze and Xen 4 in Dom0. DomU''s are Squeeze too. I thought this was a hardware problem but then I saw this (kind of old) messages: >> >> http://lists.xen.org/archives/html/xen-users/2011-02/msg00609.html >> http://bugs.debian.org/cgi-bin/bugreport.cgi?bug=599161 >> >> After reading all available info I still don''t know if the bug has been resolved in later builds of either the kernel or xen on Squeeze. Has anyone got new information on this? >> >> I see that the clocksource=pit on the grub command line does not seem to fix the problem. Any other hints? Change to Ubuntu server? Build my own xen binary? Use Debian''s backports to change the kernel version? >> >> This is kind of a big problem for me since the error kills ntpd on DomU''s and then Dovecot''s processes stop themselves when time is corrected. Scripts to workaround this are fine but I would rather much prefer for the clock not to jump ahead. >> >> Thanks a lot, >> >> Aleix. >> _______________________________________________ >> Xen-users mailing list >> Xen-users@lists.xen.org >> http://lists.xen.org/xen-users >> > > ------------------------------------------------------------------ > Steven C. Timm, Ph.D (630) 840-8525 > timm@fnal.gov http://home.fnal.gov/~timm/ > Fermilab Computing Division, Scientific Computing Facilities, > Grid Facilities Department, FermiGrid Services Group, Group Leader. > Lead of FermiCloud project.
Carlos Ribas
2012-Aug-27 16:16 UTC
Re: Clock problems on Squeeze/Xen4 (clock jumps 50 minutes ahead)
My Dom0 is Debian Squeeze with "xen-linux-system-2.6.32-5-xen-amd64" and DomU''s are Squeeze too and I don''t see the problem you said in my environment. To me this look like hardware problem. Regards, ------------------------------- Carlos Eduardo Ribas 2012/8/27 Aleix Dorca Josa <adorca@uda.ad>> Nope, all is 64 bits (Dom0 and DomU''s). > > Aleix. > > El 27/08/2012, a las 16:05, Steven Timm escribió: > > > Are you running 32-bit domU''s or 64-bit? I have seen the problem many > times under xen 3.1.2 but only with 32-bit guests on 64-bit hosts. > > > > Steve timm > > > > > > > > On Mon, 27 Aug 2012, Aleix Dorca Josa wrote: > > > >> Hi all, > >> > >> My xen farm is plagued with this ''Jump 50 minutes ahead'' problem. All > my machines run Debian Squeeze and Xen 4 in Dom0. DomU''s are Squeeze too. I > thought this was a hardware problem but then I saw this (kind of old) > messages: > >> > >> http://lists.xen.org/archives/html/xen-users/2011-02/msg00609.html > >> http://bugs.debian.org/cgi-bin/bugreport.cgi?bug=599161 > >> > >> After reading all available info I still don''t know if the bug has been > resolved in later builds of either the kernel or xen on Squeeze. Has anyone > got new information on this? > >> > >> I see that the clocksource=pit on the grub command line does not seem > to fix the problem. Any other hints? Change to Ubuntu server? Build my own > xen binary? Use Debian''s backports to change the kernel version? > >> > >> This is kind of a big problem for me since the error kills ntpd on > DomU''s and then Dovecot''s processes stop themselves when time is corrected. > Scripts to workaround this are fine but I would rather much prefer for the > clock not to jump ahead. > >> > >> Thanks a lot, > >> > >> Aleix. > >> _______________________________________________ > >> Xen-users mailing list > >> Xen-users@lists.xen.org > >> http://lists.xen.org/xen-users > >> > > > > ------------------------------------------------------------------ > > Steven C. Timm, Ph.D (630) 840-8525 > > timm@fnal.gov http://home.fnal.gov/~timm/ > > Fermilab Computing Division, Scientific Computing Facilities, > > Grid Facilities Department, FermiGrid Services Group, Group Leader. > > Lead of FermiCloud project. > > > _______________________________________________ > Xen-users mailing list > Xen-users@lists.xen.org > http://lists.xen.org/xen-users >_______________________________________________ Xen-users mailing list Xen-users@lists.xen.org http://lists.xen.org/xen-users
Aleix Dorca Josa
2012-Aug-27 16:59 UTC
Re: Clock problems on Squeeze/Xen4 (clock jumps 50 minutes ahead)
Yeah, that''s what I first thought too... but it happens at random on different machines, all using the same kernel/xen combinations but it is also true that all are using the same hardware (PowerEdge 1950III). Using xen3 I also had problems with the ''Time went backwards'' issues but changing the clocksource to jiffies solved everything. It is also true that I had not seen the issue in Xen4 for a long time since using ntpdate it corrected itself without affecting services. It wasn''t until I started using Dovecot and ntp that I had real problems. Tomorrow I''ll install Ubuntu 12.04.1 LTS on a server and move some DomU''s there. We''ll see how it works. Changing the hardware now is not possible I''m afraid. I could also try to update firmware''s and stuff on the PowerEdge, anyone done this? Thanks for your comment Carlos, Aleix. El 27/08/2012, a les 18:16, Carlos Ribas <carlos@ansp.br> va escriure:> My Dom0 is Debian Squeeze with "xen-linux-system-2.6.32-5-xen-amd64" and DomU''s are Squeeze too and I don''t see the problem you said in my environment. > > To me this look like hardware problem. > > Regards, > > ------------------------------- > Carlos Eduardo Ribas > > 2012/8/27 Aleix Dorca Josa <adorca@uda.ad> > Nope, all is 64 bits (Dom0 and DomU''s). > > Aleix. > > El 27/08/2012, a las 16:05, Steven Timm escribió: > > > Are you running 32-bit domU''s or 64-bit? I have seen the problem many times under xen 3.1.2 but only with 32-bit guests on 64-bit hosts. > > > > Steve timm > > > > > > > > On Mon, 27 Aug 2012, Aleix Dorca Josa wrote: > > > >> Hi all, > >> > >> My xen farm is plagued with this ''Jump 50 minutes ahead'' problem. All my machines run Debian Squeeze and Xen 4 in Dom0. DomU''s are Squeeze too. I thought this was a hardware problem but then I saw this (kind of old) messages: > >> > >> http://lists.xen.org/archives/html/xen-users/2011-02/msg00609.html > >> http://bugs.debian.org/cgi-bin/bugreport.cgi?bug=599161 > >> > >> After reading all available info I still don''t know if the bug has been resolved in later builds of either the kernel or xen on Squeeze. Has anyone got new information on this? > >> > >> I see that the clocksource=pit on the grub command line does not seem to fix the problem. Any other hints? Change to Ubuntu server? Build my own xen binary? Use Debian''s backports to change the kernel version? > >> > >> This is kind of a big problem for me since the error kills ntpd on DomU''s and then Dovecot''s processes stop themselves when time is corrected. Scripts to workaround this are fine but I would rather much prefer for the clock not to jump ahead. > >> > >> Thanks a lot, > >> > >> Aleix. > >> _______________________________________________ > >> Xen-users mailing list > >> Xen-users@lists.xen.org > >> http://lists.xen.org/xen-users > >> > > > > ------------------------------------------------------------------ > > Steven C. Timm, Ph.D (630) 840-8525 > > timm@fnal.gov http://home.fnal.gov/~timm/ > > Fermilab Computing Division, Scientific Computing Facilities, > > Grid Facilities Department, FermiGrid Services Group, Group Leader. > > Lead of FermiCloud project. > > > _______________________________________________ > Xen-users mailing list > Xen-users@lists.xen.org > http://lists.xen.org/xen-users >
Steve Dawson
2012-Aug-27 18:04 UTC
Re: Clock problems on Squeeze/Xen4 (clock jumps 50 minutes ahead)
On 08/27/2012 02:29 AM, Aleix Dorca Josa wrote:> Hi all, > > My xen farm is plagued with this ''Jump 50 minutes ahead'' problem. All my machines run Debian Squeeze and Xen 4 in Dom0. DomU''s are Squeeze too. I thought this was a hardware problem but then I saw this (kind of old) messages: > > http://lists.xen.org/archives/html/xen-users/2011-02/msg00609.html > http://bugs.debian.org/cgi-bin/bugreport.cgi?bug=599161 > > After reading all available info I still don''t know if the bug has been resolved in later builds of either the kernel or xen on Squeeze. Has anyone got new information on this? > > I see that the clocksource=pit on the grub command line does not seem to fix the problem. Any other hints? Change to Ubuntu server? Build my own xen binary? Use Debian''s backports to change the kernel version? > > This is kind of a big problem for me since the error kills ntpd on DomU''s and then Dovecot''s processes stop themselves when time is corrected. Scripts to workaround this are fine but I would rather much prefer for the clock not to jump ahead. > > Thanks a lot, > > Aleix. > _______________________________________________ > Xen-users mailing list > Xen-users@lists.xen.org > http://lists.xen.org/xen-users >We had this problem too, from what we could tell it was a hardware issue on some of our servers, specifically HP DL380 G5 servers, we have G6''s and G7''s as well, they don''t seem to be affected. The fix was to boot the G5 servers with clocksource=pit and cpuidle=0 on the xen command line. Example grub 1 entry below. Maybe the cpuidle switch is needed too. title Xen-4.1.1 root (hd0,0) kernel /boot/xen-4.1.1.gz dom0_mem=2048M dom0_max_vcpus=1 dom0_vcpus_pin clocksource=pit cpuidle=0 module /boot/vmlinuz-2.6.32.41 ro root=/dev/cciss/c0d0p1 max_loop=32 HTH. Steve.
Casey DeLorme
2012-Aug-27 18:54 UTC
Re: Clock problems on Squeeze/Xen4 (clock jumps 50 minutes ahead)
I had severe clock drift in a Windows HVM, I was able to fix it using ntp on the host (debian) and the "localtime=1" flag in my HVM Configuration. I am using Xen 4.2 and the xl toolstack, so I do not know if it is an option in previous versions of Xen or other toolstacks (if you are using xm for example). Easy enough to add to a configuration though, I would give it a try. On Mon, Aug 27, 2012 at 4:38 AM, Aleix Dorca Josa <adorca@uda.ad> wrote:> Hi Stephan, > > thanks for your suggestion but unfortunately what you propose only works > in Xen 3. I am using Xen 4. > > cheers, > > Aleix. > > El 27/08/2012, a les 10:30, Stephan Seitz <s.seitz@netzhaut.de> va > escriure: > > > > > > > Am Montag, den 27.08.2012, 08:29 +0200 schrieb Aleix Dorca Josa: > > > >> I see that the clocksource=pit on the grub command line does not seem > to fix the problem. Any other hints? Change to Ubuntu server? Build my own > xen binary? Use Debian''s backports to change the kernel version? > >> > >> This is kind of a big problem for me since the error kills ntpd on > DomU''s and then Dovecot''s processes stop themselves when time is corrected. > Scripts to workaround this are fine but I would rather much prefer for the > clock not to jump ahead. > >> > > > > Hi, > > > > as a quick fix, you could handle your domU''s independent from xen: > > > > echo "jiffies" > > /sys/devices/system/clocksource/clocksource0/current_clocksource > > echo "1" > /proc/sys/xen/independent_wallclock > > > > take a look at > /sys/devices/system/clocksource/clocksource0/available_clocksource to see > which clock sources are possible. > > > > cheers, > > > > Stephan > > > _______________________________________________ > Xen-users mailing list > Xen-users@lists.xen.org > http://lists.xen.org/xen-users >_______________________________________________ Xen-users mailing list Xen-users@lists.xen.org http://lists.xen.org/xen-users
MAYAP KAMGA Christine larissa
2012-Aug-27 20:54 UTC
Re: Problem with Xen-4.1.3 and Linux-3.1.0-rc9+ and Xenpm
Hello
After compiling my own Dom0 (Linux 3.1.0-rc9+) with Xen-4.1.3; i''m
unable to access the cpufreq-parameter.
# xenpm get-cpufreq-para
[CPU0] failed to get cpufreq parameter
[CPU1] failed to get cpufreq parameter
[CPU2] failed to get cpufreq parameter
[CPU3] failed to get cpufreq parameter
Though:
xenpm get-cpuidle
Max C-state: C7
cpu id : 0
total C-states : 2
idle time(ms) : 2241054
C0 : transition [00000000000000000001]
residency [00000000000000009733 ms]
C1 : transition [00000000000000000001]
residency [00000000000002241054 ms]
pc3 : [00000000000000000000 ms]
pc6 : [00000000000000000000 ms]
pc7 : [00000000000000000000 ms]
cc3 : [00000000000000000000 ms]
cc6 : [00000000000000000000 ms]
cpu id : 1
total C-states : 2
idle time(ms) : 2247721
C0 : transition [00000000000000000001]
residency [00000000000000003066 ms]
C1 : transition [00000000000000000001]
residency [00000000000002247721 ms]
pc3 : [00000000000000000000 ms]
pc6 : [00000000000000000000 ms]
pc7 : [00000000000000000000 ms]
cc3 : [00000000000000000000 ms]
cc6 : [00000000000000000000 ms]
.........
.........
#cat /boot/config-3.1.0-rc9+ | grep CPU_FREQ
CONFIG_CPU_FREQ=y
CONFIG_CPU_FREQ_TABLE=y
CONFIG_CPU_FREQ_STAT=y
# CONFIG_CPU_FREQ_STAT_DETAILS is not set
# CONFIG_CPU_FREQ_DEFAULT_GOV_PERFORMANCE is not set
# CONFIG_CPU_FREQ_DEFAULT_GOV_USERSPACE is not set
CONFIG_CPU_FREQ_DEFAULT_GOV_ONDEMAND=y
# CONFIG_CPU_FREQ_DEFAULT_GOV_CONSERVATIVE is not set
CONFIG_CPU_FREQ_GOV_PERFORMANCE=y
CONFIG_CPU_FREQ_GOV_POWERSAVE=y
CONFIG_CPU_FREQ_GOV_USERSPACE=y
CONFIG_CPU_FREQ_GOV_ONDEMAND=y
CONFIG_CPU_FREQ_GOV_CONSERVATIVE=y
#xl list
Name ID Mem
VCPUs State Time(s)
Domain-0 0 976 4 r-----
20.7
Furthermore, i''m able to access the frequency scaling information when
booting only on Linux-3.1.0-rc9+ ( i means, whithout the hypervisor).
What can be the problem?
Regards
Ian Campbell
2012-Aug-28 08:56 UTC
Re: Clock problems on Squeeze/Xen4 (clock jumps 50 minutes ahead)
On Mon, 2012-08-27 at 09:38 +0100, Aleix Dorca Josa wrote:> Hi Stephan, > > thanks for your suggestion but unfortunately what you propose only works in Xen 3. I am using Xen 4.What makes you say that? independent wallclock is a feature of the guest kernel not the hypervisor. Older kernels might have /proc/sys/xen/independent_wallclock instead, or I think you can specify it on the guest kernel command line (you''ll have to google the syntax, I don''t know it offhand). Ian.> > cheers, > > Aleix. > > El 27/08/2012, a les 10:30, Stephan Seitz <s.seitz@netzhaut.de> va escriure: > > > > > > > Am Montag, den 27.08.2012, 08:29 +0200 schrieb Aleix Dorca Josa: > > > >> I see that the clocksource=pit on the grub command line does not seem to fix the problem. Any other hints? Change to Ubuntu server? Build my own xen binary? Use Debian''s backports to change the kernel version? > >> > >> This is kind of a big problem for me since the error kills ntpd on DomU''s and then Dovecot''s processes stop themselves when time is corrected. Scripts to workaround this are fine but I would rather much prefer for the clock not to jump ahead. > >> > > > > Hi, > > > > as a quick fix, you could handle your domU''s independent from xen: > > > > echo "jiffies" > /sys/devices/system/clocksource/clocksource0/current_clocksource > > echo "1" > /proc/sys/xen/independent_wallclock > > > > take a look at /sys/devices/system/clocksource/clocksource0/available_clocksource to see which clock sources are possible. > > > > cheers, > > > > Stephan > > > _______________________________________________ > Xen-users mailing list > Xen-users@lists.xen.org > http://lists.xen.org/xen-users
Aleix Dorca Josa
2012-Aug-28 09:48 UTC
Re: Clock problems on Squeeze/Xen4 (clock jumps 50 minutes ahead)
Hi Stephan, maybe I misunderstood but I was kind of sure that the ''jiffies'' thing could not be used with Xen4. I''ve looked at my domU and the jiffies clocksource is not available, only xen (the current one) and tsc. Also on a Squeeze DomU there''s no independent_wallclock on proc fs. Maybe it has another name? I''ve used the timer_mode on some HVM domU''s but I don''t really remember why... I think it has nothing to do. Thanks, Aleix. El 28/08/2012, a les 10:56, Ian Campbell <Ian.Campbell@citrix.com> va escriure:> On Mon, 2012-08-27 at 09:38 +0100, Aleix Dorca Josa wrote: >> Hi Stephan, >> >> thanks for your suggestion but unfortunately what you propose only works in Xen 3. I am using Xen 4. > > What makes you say that? independent wallclock is a feature of the guest > kernel not the hypervisor. > > Older kernels might have /proc/sys/xen/independent_wallclock instead, or > I think you can specify it on the guest kernel command line (you''ll have > to google the syntax, I don''t know it offhand). > > Ian. > >> >> cheers, >> >> Aleix. >> >> El 27/08/2012, a les 10:30, Stephan Seitz <s.seitz@netzhaut.de> va escriure: >> >>> >>> >>> Am Montag, den 27.08.2012, 08:29 +0200 schrieb Aleix Dorca Josa: >>> >>>> I see that the clocksource=pit on the grub command line does not seem to fix the problem. Any other hints? Change to Ubuntu server? Build my own xen binary? Use Debian''s backports to change the kernel version? >>>> >>>> This is kind of a big problem for me since the error kills ntpd on DomU''s and then Dovecot''s processes stop themselves when time is corrected. Scripts to workaround this are fine but I would rather much prefer for the clock not to jump ahead. >>>> >>> >>> Hi, >>> >>> as a quick fix, you could handle your domU''s independent from xen: >>> >>> echo "jiffies" > /sys/devices/system/clocksource/clocksource0/current_clocksource >>> echo "1" > /proc/sys/xen/independent_wallclock >>> >>> take a look at /sys/devices/system/clocksource/clocksource0/available_clocksource to see which clock sources are possible. >>> >>> cheers, >>> >>> Stephan >> >> >> _______________________________________________ >> Xen-users mailing list >> Xen-users@lists.xen.org >> http://lists.xen.org/xen-users > >
Stephan Seitz
2012-Aug-28 10:07 UTC
Re: Clock problems on Squeeze/Xen4 (clock jumps 50 minutes ahead)
Hi, as Ian already said, clocks are handled by the particular kernel(s). On a recent CentOS 6, I''m using echo "tsc"> /sys/devices/system/clocksource/clocksource0/current_clocksourceThat VM is currently running on top of a Xen 4.1.3. Oh, and it''s running ntpd without problems. I''ve just compared against the settings of an Ubuntu 12.04 on top of a Xen 4.0.2 and noticed, it''s exactly the same. I think the independent_wallclock key left the procfs a while ago. I''m not quite sure, but I assume the result of having an independent clock can be achieved by only setting current_clocksource . I don''t have a Debian Squeeze VM at hand, so I can''t tell which procfs / sysfs keys are available. Cheers, Stephan Am Dienstag, den 28.08.2012, 11:48 +0200 schrieb Aleix Dorca Josa:> Hi Stephan, > > maybe I misunderstood but I was kind of sure that the ''jiffies'' thing could not be used with Xen4. > > I''ve looked at my domU and the jiffies clocksource is not available, only xen (the current one) and tsc. > Also on a Squeeze DomU there''s no independent_wallclock on proc fs. Maybe it has another name? > > I''ve used the timer_mode on some HVM domU''s but I don''t really remember why... I think it has nothing to do. > > Thanks, > > Aleix. > > El 28/08/2012, a les 10:56, Ian Campbell <Ian.Campbell@citrix.com> va escriure: > > > On Mon, 2012-08-27 at 09:38 +0100, Aleix Dorca Josa wrote: > >> Hi Stephan, > >> > >> thanks for your suggestion but unfortunately what you propose only works in Xen 3. I am using Xen 4. > > > > What makes you say that? independent wallclock is a feature of the guest > > kernel not the hypervisor. > > > > Older kernels might have /proc/sys/xen/independent_wallclock instead, or > > I think you can specify it on the guest kernel command line (you''ll have > > to google the syntax, I don''t know it offhand). > > > > Ian. > > > >> > >> cheers, > >> > >> Aleix. > >> > >> El 27/08/2012, a les 10:30, Stephan Seitz <s.seitz@netzhaut.de> va escriure: > >> > >>> > >>> > >>> Am Montag, den 27.08.2012, 08:29 +0200 schrieb Aleix Dorca Josa: > >>> > >>>> I see that the clocksource=pit on the grub command line does not seem to fix the problem. Any other hints? Change to Ubuntu server? Build my own xen binary? Use Debian''s backports to change the kernel version? > >>>> > >>>> This is kind of a big problem for me since the error kills ntpd on DomU''s and then Dovecot''s processes stop themselves when time is corrected. Scripts to workaround this are fine but I would rather much prefer for the clock not to jump ahead. > >>>> > >>> > >>> Hi, > >>> > >>> as a quick fix, you could handle your domU''s independent from xen: > >>> > >>> echo "jiffies" > /sys/devices/system/clocksource/clocksource0/current_clocksource > >>> echo "1" > /proc/sys/xen/independent_wallclock > >>> > >>> take a look at /sys/devices/system/clocksource/clocksource0/available_clocksource to see which clock sources are possible. > >>> > >>> cheers, > >>> > >>> Stephan > >> > >> > >> _______________________________________________ > >> Xen-users mailing list > >> Xen-users@lists.xen.org > >> http://lists.xen.org/xen-users > > > > >_______________________________________________ Xen-users mailing list Xen-users@lists.xen.org http://lists.xen.org/xen-users
Aleix Dorca Josa
2012-Aug-28 10:22 UTC
Re: Clock problems on Squeeze/Xen4 (clock jumps 50 minutes ahead)
Stephan, Just to clarify things the ''echo'' command, where have you executed it, on the Dom0 or on the DomU, or on both? On Ubuntu 12.04.1 running Xen 4.2.1 the setting are the same. It seems to be pretty straightforward. I''m going to try it right away. Thanks a lot, Aleix. El 28/08/2012, a les 12:07, Stephan Seitz <s.seitz@netzhaut.de> va escriure:> Hi, > > as Ian already said, clocks are handled by the particular kernel(s). > > On a recent CentOS 6, I''m using > > echo "tsc" > /sys/devices/system/clocksource/clocksource0/current_clocksource > > That VM is currently running on top of a Xen 4.1.3. > Oh, and it''s running ntpd without problems. > > I''ve just compared against the settings of an Ubuntu 12.04 on top of a Xen 4.0.2 > and noticed, it''s exactly the same. > > I think the independent_wallclock key left the procfs a while ago. I''m not quite > sure, but I assume the result of having an independent clock can be achieved > by only setting current_clocksource . > > I don''t have a Debian Squeeze VM at hand, so I can''t tell which procfs / sysfs keys > are available. > > Cheers, > > Stephan > > > > > > > > Am Dienstag, den 28.08.2012, 11:48 +0200 schrieb Aleix Dorca Josa: >> Hi Stephan, >> >> maybe I misunderstood but I was kind of sure that the ''jiffies'' thing could not be used with Xen4. >> >> I''ve looked at my domU and the jiffies clocksource is not available, only xen (the current one) and tsc. >> Also on a Squeeze DomU there''s no independent_wallclock on proc fs. Maybe it has another name? >> >> I''ve used the timer_mode on some HVM domU''s but I don''t really remember why... I think it has nothing to do. >> >> Thanks, >> >> Aleix. >> >> El 28/08/2012, a les 10:56, Ian Campbell < >> Ian.Campbell@citrix.com >> > va escriure: >> >> > On Mon, 2012-08-27 at 09:38 +0100, Aleix Dorca Josa wrote: >> >> Hi Stephan, >> >> >> >> thanks for your suggestion but unfortunately what you propose only works in Xen 3. I am using Xen 4. >> > >> > What makes you say that? independent wallclock is a feature of the guest >> > kernel not the hypervisor. >> > >> > Older kernels might have /proc/sys/xen/independent_wallclock instead, or >> > I think you can specify it on the guest kernel command line (you''ll have >> > to google the syntax, I don''t know it offhand). >> > >> > Ian. >> > >> >> >> >> cheers, >> >> >> >> Aleix. >> >> >> >> El 27/08/2012, a les 10:30, Stephan Seitz < >> s.seitz@netzhaut.de >> > va escriure: >> >> >> >>> >> >>> >> >>> Am Montag, den 27.08.2012, 08:29 +0200 schrieb Aleix Dorca Josa: >> >>> >> >>>> I see that the clocksource=pit on the grub command line does not seem to fix the problem. Any other hints? Change to Ubuntu server? Build my own xen binary? Use Debian''s backports to change the kernel version? >> >>>> >> >>>> This is kind of a big problem for me since the error kills ntpd on DomU''s and then Dovecot''s processes stop themselves when time is corrected. Scripts to workaround this are fine but I would rather much prefer for the clock not to jump ahead. >> >>>> >> >>> >> >>> Hi, >> >>> >> >>> as a quick fix, you could handle your domU''s independent from xen: >> >>> >> >>> echo "jiffies" > /sys/devices/system/clocksource/clocksource0/current_clocksource >> >>> echo "1" > /proc/sys/xen/independent_wallclock >> >>> >> >>> take a look at /sys/devices/system/clocksource/clocksource0/available_clocksource to see which clock sources are possible. >> >>> >> >>> cheers, >> >>> >> >>> Stephan >> >> >> >> >> >> _______________________________________________ >> >> Xen-users mailing list >> >> >> Xen-users@lists.xen.org >> >> >> >> http://lists.xen.org/xen-users >> >> > >> > >> >>
Christine MAYAP
2012-Aug-28 13:42 UTC
Re: Problem with Xen-4.1.3 and Linux-3.1.0-rc9+ and Xenpm
Hello Please, is there somebody concerns about this problem(see below)?? On 27/08/2012 22:54, MAYAP KAMGA Christine larissa wrote:> Hello > > After compiling my own Dom0 (Linux 3.1.0-rc9+) with Xen-4.1.3; i''m > unable to access the cpufreq-parameter. > # xenpm get-cpufreq-para > [CPU0] failed to get cpufreq parameter > [CPU1] failed to get cpufreq parameter > [CPU2] failed to get cpufreq parameter > [CPU3] failed to get cpufreq parameter > > Though: > xenpm get-cpuidle > Max C-state: C7 > > cpu id : 0 > total C-states : 2 > idle time(ms) : 2241054 > C0 : transition [00000000000000000001] > residency [00000000000000009733 ms] > C1 : transition [00000000000000000001] > residency [00000000000002241054 ms] > pc3 : [00000000000000000000 ms] > pc6 : [00000000000000000000 ms] > pc7 : [00000000000000000000 ms] > cc3 : [00000000000000000000 ms] > cc6 : [00000000000000000000 ms] > > cpu id : 1 > total C-states : 2 > idle time(ms) : 2247721 > C0 : transition [00000000000000000001] > residency [00000000000000003066 ms] > C1 : transition [00000000000000000001] > residency [00000000000002247721 ms] > pc3 : [00000000000000000000 ms] > pc6 : [00000000000000000000 ms] > pc7 : [00000000000000000000 ms] > cc3 : [00000000000000000000 ms] > cc6 : [00000000000000000000 ms] > ......... > ......... > > #cat /boot/config-3.1.0-rc9+ | grep CPU_FREQ > CONFIG_CPU_FREQ=y > CONFIG_CPU_FREQ_TABLE=y > CONFIG_CPU_FREQ_STAT=y > # CONFIG_CPU_FREQ_STAT_DETAILS is not set > # CONFIG_CPU_FREQ_DEFAULT_GOV_PERFORMANCE is not set > # CONFIG_CPU_FREQ_DEFAULT_GOV_USERSPACE is not set > CONFIG_CPU_FREQ_DEFAULT_GOV_ONDEMAND=y > # CONFIG_CPU_FREQ_DEFAULT_GOV_CONSERVATIVE is not set > CONFIG_CPU_FREQ_GOV_PERFORMANCE=y > CONFIG_CPU_FREQ_GOV_POWERSAVE=y > CONFIG_CPU_FREQ_GOV_USERSPACE=y > CONFIG_CPU_FREQ_GOV_ONDEMAND=y > CONFIG_CPU_FREQ_GOV_CONSERVATIVE=y > > #xl list > Name ID Mem VCPUs State > Time(s) > Domain-0 0 976 4 > r----- 20.7 > > > > Furthermore, i''m able to access the frequency scaling information when > booting only on Linux-3.1.0-rc9+ ( i means, whithout the hypervisor). > What can be the problem? > > > Regards-- MAYAP Christine IRIT/ENSEEIHT 2 rue Charles Camichel - BP 7122 31071 Toulouse cedex 7
Pasi Kärkkäinen
2012-Aug-28 18:41 UTC
Re: Problem with Xen-4.1.3 and Linux-3.1.0-rc9+ and Xenpm
On Mon, Aug 27, 2012 at 10:54:52PM +0200, MAYAP KAMGA Christine larissa wrote:> Hello > > After compiling my own Dom0 (Linux 3.1.0-rc9+) with Xen-4.1.3; i''m > unable to access the cpufreq-parameter. > # xenpm get-cpufreq-para > [CPU0] failed to get cpufreq parameter > [CPU1] failed to get cpufreq parameter > [CPU2] failed to get cpufreq parameter > [CPU3] failed to get cpufreq parameter >Linux 3.1 does not have xen-acpi-processor driver. Also why do you use such a weird version 3.1.0-rc9 ? Try latest 3.4.x kernel, it has xen-acpi-processor driver, which is required for cpufreq stuff. -- Pasi> Though: > xenpm get-cpuidle > Max C-state: C7 > > cpu id : 0 > total C-states : 2 > idle time(ms) : 2241054 > C0 : transition [00000000000000000001] > residency [00000000000000009733 ms] > C1 : transition [00000000000000000001] > residency [00000000000002241054 ms] > pc3 : [00000000000000000000 ms] > pc6 : [00000000000000000000 ms] > pc7 : [00000000000000000000 ms] > cc3 : [00000000000000000000 ms] > cc6 : [00000000000000000000 ms] > > cpu id : 1 > total C-states : 2 > idle time(ms) : 2247721 > C0 : transition [00000000000000000001] > residency [00000000000000003066 ms] > C1 : transition [00000000000000000001] > residency [00000000000002247721 ms] > pc3 : [00000000000000000000 ms] > pc6 : [00000000000000000000 ms] > pc7 : [00000000000000000000 ms] > cc3 : [00000000000000000000 ms] > cc6 : [00000000000000000000 ms] > ......... > ......... > > #cat /boot/config-3.1.0-rc9+ | grep CPU_FREQ > CONFIG_CPU_FREQ=y > CONFIG_CPU_FREQ_TABLE=y > CONFIG_CPU_FREQ_STAT=y > # CONFIG_CPU_FREQ_STAT_DETAILS is not set > # CONFIG_CPU_FREQ_DEFAULT_GOV_PERFORMANCE is not set > # CONFIG_CPU_FREQ_DEFAULT_GOV_USERSPACE is not set > CONFIG_CPU_FREQ_DEFAULT_GOV_ONDEMAND=y > # CONFIG_CPU_FREQ_DEFAULT_GOV_CONSERVATIVE is not set > CONFIG_CPU_FREQ_GOV_PERFORMANCE=y > CONFIG_CPU_FREQ_GOV_POWERSAVE=y > CONFIG_CPU_FREQ_GOV_USERSPACE=y > CONFIG_CPU_FREQ_GOV_ONDEMAND=y > CONFIG_CPU_FREQ_GOV_CONSERVATIVE=y > > #xl list > Name ID Mem > VCPUs State Time(s) > Domain-0 0 976 4 > r----- 20.7 > > > > Furthermore, i''m able to access the frequency scaling information > when booting only on Linux-3.1.0-rc9+ ( i means, whithout the > hypervisor). > What can be the problem? > > > Regards > > > _______________________________________________ > Xen-users mailing list > Xen-users@lists.xen.org > http://lists.xen.org/xen-users
Mauro
2012-Sep-30 20:13 UTC
Re: Clock problems on Squeeze/Xen4 (clock jumps 50 minutes ahead)
On 27 August 2012 20:04, Steve Dawson <sdawson@pa.metrocast.net> wrote:> We had this problem too, from what we could tell it was a hardware issue on some of our servers, specifically HP DL380 G5 servers, we have G6''s and G7''s as well, they don''t seem to be affected. > > The fix was to boot the G5 servers with clocksource=pit and cpuidle=0 on the xen command line. Example grub 1 entry below. Maybe the cpuidle switch is needed too. > > title Xen-4.1.1 > root (hd0,0) > kernel /boot/xen-4.1.1.gz dom0_mem=2048M dom0_max_vcpus=1 dom0_vcpus_pin clocksource=pit cpuidle=0 > module /boot/vmlinuz-2.6.32.41 ro root=/dev/cciss/c0d0p1 max_loop=32Hello, I have exactly the same problem on two HP Proliant DL580 G5 servers with debian squeeze. It''s very very annoying and I am also angry for this bug. Is it a xen bug or a hardware problem? I''ve tried clocksource=pit but I''ve not resolved. Have you solved adding cpuidle=0?