Hello, I’m seeing clock drift in my application’s container. It’s at about -1.15s off, and seems to survive app restarts (guess it’s the host’s clock that’s drifted?). Do you recommend running ntpd/chrony in containers? Or some other fix in the guest?
The clock shouldn’t drift, which region is this happening in?
That’s interesting. Our init updates the clock in the VM every 15s. Firecracker times drift, but I’m surprised it’s that much: firecracker/FAQ.md at main · firecracker-microvm/firecracker · GitHub
How are you detecting the drift, is is something we can replicate?
Sure, in an alpine container, i run
echo 'server 0.us.pool.ntp.org' > /etc/ntp.conf ; ntpd -n and see “ntpd: setting time to 2021-11-01 15:31:26.319249 (offset -1.159240s)” in the output.
chronyd also prints “System’s initial offset : 1.158265 seconds fast of true (slew)” when i run it manually.
Ok thanks, give us a bit to investigate.
@benburkert how does it look now?
looks good now, thanks!