Unruh wrote: > A question has come up on another newgroup whethr or not the time can be > disciplined via ntp (or chrony) on a virtual server. I would suspect not > (imagine the mess if both ntp on one virtual server and chrony on another > were both trying to discipline the same hardware clock, and I cannot see > how the virtual OS could keep track of its own separate clock since it only > gets some subset of the time slices on the real system) but perhaps I am > wrong. Can a virtual server discipline its own clock? If not, how is the > adjtimex system call trapped? Surely each virtual machine cannot influence > the real system, which is what would happen if the adjtimex call on each > virtual machine were passed to the adjtimex system call on the real > machine. > You should discipline the host and run vmware tools or equivalent on the guests. You should not run anything with tight timing requirements on a guest, as the guest's time will have large short term variations because the host will dish out time roughly in proportion to the CPU it gives.
A link to the VMWare paper on this was posted last week. _______________________________________________ questions mailing list [email protected] https://lists.ntp.org/mailman/listinfo/questions
