Rudder

Imagine you have nodes with a run interval of interval=60min.

The node agent run at t0.

At t1 = t0+a few minutes, we change the run interval properties for the node to interval=5min.

The agent won't run until t0+60min, so it won't see the run interval modification until that time. But for the server, the run interval is now 5min, so it does expect a run before t0+5min(+5min of grace period).

So, from t0+10 until t0+60min, the node will be reported in no data, which is actually true - but it is not due to a problem.

Something alike may happens when switching from "changes only" to "full compliance" mode, if the agent didn't sent reports since several run due to heartbeat.

I see three ways of correcting/preventing that:

- 1/ store in the server for each config version what were the compliance&run interval parameters (can be done with some hours of work and a change in database schema);
- 2/ force agent run on such changes (what suppose to be able to contact them all and is the opposite of much of what we are doing now)
- 3/ add a comment that "when reducing notably the run interval length, "no data" may be reported for agent until their next run" (well, something like that but actually well written)