Is there any other side effects when the system clock differs on each node

[rinsozheng]

Please read https://github.com/coreos/etcd/blob/master/Documentation/reporting_bugs.md.

We are running 3 ETCD nodes, the time on each node synced with NTP server, while for some time, the operation system clock differs with each other(because of NTP network issue on some nodes), it keeps getting the below information from the logs:

2018-05-23 17:00:29.475936 W | rafthttp: the clock difference against peer 5bd0394181f135dd is too high [10.449391606s > 1s]
2018-05-23 17:00:59.476151 W | rafthttp: the clock difference against peer 5bd0394181f135dd is too high [10.449443811s > 1s]
2018-05-23 17:01:29.476315 W | rafthttp: the clock difference against peer 5bd0394181f135dd is too high [10.449597395s > 1s]
2018-05-23 17:01:59.476514 W | rafthttp: the clock difference against peer 5bd0394181f135dd is too high [10.449581152s > 1s]

Right now, we can only see the log warnings above, there is no leader election/data replication issues happen so far.

My question here is: Is there any other side effects if the system clock differs, can we just ignore such information?

./etcd --version

etcd Version: 3.2.7
Git SHA: bb66589
Go Version: go1.8.3
Go OS/Arch: linux/amd64

[xiang90]

you should fix your clock. or TTL or lease might be effected.

[rinsozheng]

Thanks!

[pjhdld]

你应该修理时钟。或TTL或租约。

@xiang90
I want to know what is the different impact of each node clock, TTL or lease? Or will it cause other problems?