horazont
commented
4 years ago The reason is that no periodic sync is even implemented, hah.
In fact we only sync on changes which could affect our config, which is often enough in a busy cluster (endpoints and node objects updating). In a more idle cluster, you don’t see such changes very often.
Of course, a periodic sync is desirable for ensuring consistency even if agents lose connectivity for a while or are restarted or added or whatever, so it should be added. Let us fight the eternal fight against entropy
We’re seeing in our clusters that the periodic safeguard sync isn’t working all the time. Syncs required due to changed configuration are executed properly though.