Am 8. März 2019 20:29:36 MEZ schrieb Dave Voutila <d...@sisu.io>: >On Fri, Mar 8, 2019 at 1:55 PM Joe M <joe9m...@gmail.com> wrote: >> >> I have the same issue and have been using this driver. It sets the >> correct time every 5 seconds. For this purpose, this solution is a >> hack, but, I could not figure out a better solution. >> >> https://github.com/voutilad/virtio_vmmci/issues/1 > >As the author of the driver, let me clarify that the behavior now >mimics the native vmmci(4) behavior in that it only syncs the clock >when told to by the host system, which corresponds to the virtual >mc146818 detected its own clock drift (see vmd(8) source code). In >practice, this usually happens when the host system comes out of >suspension/hibernation. > >The driver is not meant to resolve major issues like chronic clock >drift that even ntp can't fix.
The drift is not my primary problem. Yes, the drift is still large but it really solves the problem that "containerd" does not eat 100% CPU. It was unusable before. But I'll keep that in mind. Thanks Robert > >> >> Also, I noticed that vm clock would be very slow. It loses 28 minutes >> for every 30 minutes. >> >> Another thing, watch out regarding tsc=unreliable kernel command line >> option, the vm is getting unresponsive with it after an hour or so. >> > >-Dave -- sent from my mobile device
