Hi Ben, Thanks, slewing sounds like the right design. It keeps time monotonic, which was the point of 36770, and still lets workers converge back to main.
Yes please on the tentative patch. We'll test it on our lab node and watch "show clock" skew over time. One data point for sizing the slew rate: we saw about 3 s of worker skew after 33 days, and Baldur reported 27 s after 57 days. So the drift is well under 1 s/day. A correction rate of even ~100 ppm would converge quickly without noticeable jumps. Separately, would you still want the lacp_time_now() change upstream as defence in depth, so LACP doesn't depend on per-thread clocks at all? Happy to push it if so. Regards, Pieter Meyer
-=-=-=-=-=-=-=-=-=-=-=- Links: You receive all messages sent to this group. View/Reply Online (#27223): https://lists.fd.io/g/vpp-dev/message/27223 Mute This Topic: https://lists.fd.io/mt/121360743/21656 Group Owner: [email protected] Unsubscribe: https://lists.fd.io/g/vpp-dev/leave/14379924/21656/631435203/xyzzy [[email protected]] -=-=-=-=-=-=-=-=-=-=-=-
