Serverless scheduling

[innermatrix]

My goal is to use the pg-boss scheduler in serverless context. The idea is along the lines of

• Run a lambda on a native AWS timer (using EventBridge)
• Spawn a pg-boss instance in that lambda to find all scheduled tasks that need to be run
• Shut down the pg-boss instance
• Repeat

But the problem that I have with this is that currently (as I understand from reading the code) the pg-boss scheduler will only run a scheduled task if the scheduler invocation takes place <60s after the scheduled task time, and the shortest interval that EventBridge timers can use to invoke lambdas is 1min. So there's a small, but non-zero, probability that a scheduled task will be missed if the timing of the lambda invocation is such that one scheduler check takes place slightly before "scheduled time N + 60s" and the next scheduler check takes place slightly after "scheduled time N+1 + 60s".

Furthermore, maybe I don't want to pay to run a lambda every 1min — maybe I'd rather run the lambda every 10min, as long as I am okay with the implications that has on my scheduled tasks (meaning that I am okay if 5 * * * * and 6 * * * * schedules actually both run at ~the same time around 10min after the hour, and a 5,6 * * * * schedule gets debounced so it only runs once per hour at around 10min after the hour).

So, what I am wondering is

• Did I understand the behavior of the scheduler correctly wrt only running tasks that are <60s overdue?
• Are there any recommendations for how to run the scheduler in a periodic lambda context (given its restrictions of max rate of one lambda invocation per minute)
• Is there any interest in making changes to the scheduler so that when it's run less frequently than once per minute it operates in a catch-up mode rather than in a skip mode?

Thanks!