0

I’ve been reading over ntp documentation and I’m not quite sure I’m wrapping my head around ‘stepout’. We have a large environment of Linux VM’s that require ‘tinker panic 0’ to be set. ntp is configured in slew mode (required) and VM’s will routinely have clocks skewed by large values.

I’m trying to tune ntp.conf so that if the time cannot be corrected to 5 seconds within 30 minutes that the service steps the clock as to avoid manually having to manually restart the service.

I configured stepout 60, and manually changed the clock on a VM. ntp did in fact step the time but it took about 20 minutes for it to do so.

I’m not sure if I’m interpreting documentation for the service correctly and am hoping someone has any insight on optimizing ntp.conf for VM’s can offer any suggestions/insight.

Michael Moser
  • 219
  • 2
  • 4
  • 16
  • Have you considered chronyd? It should work much better than ntpd, especially with clients which are not constantly online, or powered on. – dexter Apr 05 '20 at 11:45
  • 1
    If your VMs are routinely skewed by large amounts, there's something else wrong. That really shouldn't happen under any modern hypervisor or current Linux kernel. What is the hypervisor, and what steps have you taken to try to correct the skew? – Paul Gear Apr 05 '20 at 12:09
  • Chronyd will happen whenever the powers that be make that determination unfortunately. These are running on VMWare; they are stable unless vmotioned; at which point they get skewed. – Michael Moser Apr 05 '20 at 23:25
  • Are the VMware hosts themselves in close sync? Is host time sync enabled or disabled? – Paul Gear Apr 06 '20 at 08:02
  • The hosts and BIOS settings are configured. If the clock is off by more than 5,000 MS for more than 30 minutes I need to step instead of slew. If a host migration occurs the clock gets skewed, some VMs are much larger than others. – Michael Moser Apr 09 '20 at 23:04

0 Answers0