I have an etcd cluster set up on 3 nodes running rancher OS. After about 8 hours, the etcd container on one of my hosts started warning me that the clock difference against both other nodes was too high (> 1s). The difference has been (slowly) climbing and over the course of ~4-5 hours has gone from ~1.0s to ~1.6s. Neither of the other 2 hosts report any clock synchronization issues. I’ve tried restarting the ntp containers on all hosts but the difference persists. Any idea what’s causing this drift?
I have same issue.