Scheduler Minimum Interval and Last Run

[brianghig]

The scheduler service shows some inconsistent / unexpected behavior with scheduling. Perhaps this is simply my misunderstanding, but from what I'm seeing:

1. the minimum interval (set globally in config.'scheduler.interval' with a default of 10s) disallows any scheduled services with an interval less than 10s because the global timeout only runs on this schedule
2. the timeoutHandler checks if the current timestamp is greater than the lastRun + the serviceInterval. So, if a job last ran 10 seconds ago and is scheduled for 10 second intervals, it may not run until the next iteration.
3. the lastRun value for each service is only set after the service completes. So, if a job is scheduled to run every 10 seconds, and takes 200ms to run, it would only run every 20 seconds.

Proposed changes:

• Set the global timeout interval for the timeoutHandler to the greatest common factor (so that a 10s interval doesn't artificially delay a 15s interval)
• Change the time check to a "greater than or equal to" check
• Set the lastRun value to the timestamp when the service starts. (Note: we do not run the risk of concurrent executions because of the check on the boolean flag of running that is maintained for each service.)

[brianghig]

It may be worth looking into implementing this with a third party library like https://github.com/agenda/agenda that would allow for persistence and multi-node support (e.g., only running a job on one node in a cluster by using MongoDB for locking support).