Symptoms¶
- A 3-node etcd cluster is reporting consensus failures and leader election instability.
- Application logs show:
request timestamp too far in the futureandclock skew detected. - TLS certificate validation is intermittently failing with "certificate not yet valid" errors.
- One node (
node-2) shows a system time that is 4 minutes and 37 seconds ahead of the others. - The issue started after
node-2was rebooted following a hardware maintenance event. chronyc trackingon node-2 shows the NTP source is unreachable.- The other two nodes have synchronized clocks.