https://dagster.io/ logo
#deployment-kubernetes
Title
# deployment-kubernetes
p

Philippe Laflamme

07/06/2023, 1:54 PM
In the context of running within k8s and the fact that things can restart at any time: what's the behaviour around schedules in the event that the daemon was down when a tick passed? Say, for example, I have a schedule and its next tick is at today 9:00AM, but the deamon restarts exactly around this point and "misses" the tick (say because it restarted at 9:01AM). Does scheduler notice it missed a tick or does it simply compute the next tick and waits for that one to occur? I've only been using sensors which don't have such a problem (generally), but now I have a case where I might need to use a schedule and I'm wondering how to prevent gaps...
b

ben

07/06/2023, 4:25 PM
Hi Philippe, for a “partitioned schedule”, the scheduler will “catch up” on missed ticks, up to the
max_catchup_runs
config value (default 5). Here are the docs on that config option: https://docs.dagster.io/_apidocs/schedules-sensors#dagster._core.scheduler.DagsterDaemonScheduler Otherwise, for non-partitioned schedules, the scheduler will only catch up on the last tick
🙏 1
p

Philippe Laflamme

07/06/2023, 4:47 PM
Thanks for this! One follow-up question: I'm using
end_offset = 2
on that partition definition because I need to load "tomorrow's" data. That is, every morning (say at 9AM), I want the schedule to kick in so that it launches a run for
partition_key=today + 1
Does "partitioned schedule" understand this?