On Thu, Jun 04, 2009 at 06:50:36AM -0500, Michael S. Zick wrote:
> Here is something I noticed which does not
> make any difference in my use case, but.. FYI:
> My host server is used to host guest distributions
> - - mostly for software development
> - - my host "sleeps" (suspend to ram) when I do
> For the host:
> wolf-c2q ~ # uptime
> 06:37:52 up 9 days, 14:11, 3 users, load average: 0.20, 0.06, 0.02
> Which looks about correct (it is elapsed wall time) -
> *including the wall time the system was asleep*
> For the guests:
> wolf-c2q ~ # vserver-stat
> CTX PROC VSZ RSS userTIME sysTIME UPTIME NAME
> 14156 2 8.4M 1.6M 0m00s42 0m00s49 1d08h43 deb-wolf
> 14158 2 6M 1.5M 4h16m10 24m52s83 1d08h43 gen2-32
> Which looks a little bit strange -
> even if the "asleep time" was deducted.
the uptime virtualization is implemented using an
'offset' a bias_uptime, which basically stores the
time when the guest is started and shows the time
difference between now and that stored values as
uptime inside the guest
so, as long as time progresses in a linear way,
host and guest uptime should not drift (i.e. they
should keep the same offset)
> I **know** I have had them awake more than 32 hours in the
> past 9 days (I don't know the exact time, but > 1d08h)
maybe the guests simply got restarted? looks like
they have the same uptime ...
HTH,
Herbert
> - - - -
>
> Versions:
> Kernel: 2.6.27.10-vs2.3.0.36.2
> VS-API: 0x00020304
> util-vserver: 0.30.215; Dec 17 2008, 10:37:16
>
> Mike
Received on Sun Jun 7 00:12:03 2009