Catchup & Power Events

Catchup & Power Events

Cron schedulers running on user machines have to deal with reality: laptops sleep, apps quit, networks drop. Ptah’s scheduler handles these explicitly through a catchup policy and an OS power-monitor integration.

What gets missed

Anything scheduled to fire while:

• The OS is asleep / hibernated
• Ptah isn’t running
• The job was disabled and just got re-enabled

…is “missed” and subject to the catchup policy.

Catchup policies

Set per-job:

Policy	Behaviour after the app wakes up
none	Skip everything missed. Just resume on the next normal slot
last	Fire the most recent missed slot once. Older missed slots are dropped
all	Fire every missed slot, oldest first, up to cron.catchupWindowMs (hard-capped at 24 hours)

Power-monitor integration

On wake events, Ptah:

1. Re-arms in-memory timers for every active job
2. Asks the CatchupCoordinator which jobs have nextRunAt < now
3. Replays them according to each job’s policy

The catchup window is hard-capped at 24 hours regardless of cron.catchupWindowMs. A laptop that’s been closed for a week won’t suddenly fire 168 missed hourly jobs.

Replays can surprise you

Don’t use all for jobs that produce side effects — git commits, social posts, emails, deploys. A laptop closed overnight could fire a dozen replays the moment you open the lid. last or none is almost always safer for those.

Skipped runs

Slots that hit the policy filter (or the global concurrency cap) are recorded in the run history with status skipped and an explanation. Nothing missed disappears silently — you can audit it.