UTC semantics and the fire-window
Two facts cause 90 % of "my schedule didn't fire when I expected" tickets: 1. All times are UTC. The form does not convert from your local timezone. 2. Real fire is 0–5 minutes after the planned minute. Webhooks Pro runs on Forge's 5-minute scheduled trigger.
If you internalise these two, scheduled webhooks become predictable.
Always UTC
Every minute / hour / day-of-week input on the Schedule card is UTC. There is no per-tenant timezone setting and no auto-conversion from your browser's locale.
Why UTC: it's the only timezone that doesn't shift twice a year and doesn't depend on who's looking at the form. Every other approach trades obviousness for stability, and we picked stability.
What it means in practice
You wrote | Real fire time |
|---|---|
| 09:00 UTC on weekdays |
→ in London winter (GMT) | 09:00 |
→ in London summer (BST, UTC+1) | 10:00 |
→ in Berlin winter (CET, UTC+1) | 10:00 |
→ in Berlin summer (CEST, UTC+2) | 11:00 |
→ in New York winter (EST, UTC−5) | 04:00 |
→ in New York summer (EDT, UTC−4) | 05:00 |
→ in Moscow (MSK, UTC+3, no DST) | 12:00 |
If your audience spans daylight-saving changes, you have two choices:
Pick a UTC time once. Local time will shift twice a year. Most teams accept this — they think in UTC anyway.
Update the CRON twice a year. Higher fidelity to local time, but you have to remember to do it.
A copy-paste cheat sheet for common business hours: Schedule examples.
The 0–5 minute fire-window
Forge runs the underlying scheduled trigger every 5 minutes on a fixed cadence. The exact offset is set by Atlassian per-tenant at install time:
Your tenant might tick at
:00 / :05 / :10 / :15 / …Another tenant might tick at
:02 / :07 / :12 / :17 / …
The cadence is constant for the lifetime of the install. If you re-install, the offset can shift.
On each tick, Webhooks Pro looks back over the past few minutes and fires any schedule whose CRON matched a minute in that window. So:
A webhook never fires earlier than its CRON time.
A webhook can fire up to 5 minutes after its CRON time.
Concrete trace
A schedule of 0 0 12 ? * SUN * (Sundays at 12:00 UTC) on a tenant whose Forge tick lands at :02 / :07 / :12 / …:
Sunday 11:57 UTC tick → looks back 11:52–11:57 → no match → skip
Sunday 12:02 UTC tick → looks back 11:57–12:02 → 12:00 matched → fireThe webhook fired 2 minutes late. Other tenants will see different offsets, but always inside the 0–5 minute upper bound.
Implications
Don't build a workflow that needs exactly 12:00:00 UTC. The bound is "between 12:00 and 12:05 UTC".
Don't compare wall-clock receiver time to scheduled time and alert on mismatch unless your tolerance is
≥ 5 minutes.For schedules tighter than every 5 minutes, the actual cadence collapses to "every 5 minutes" — you can ask for finer, but you'll only ever fire on the platform tick.
What if the tick is missed
It happens — Forge maintenance, region failover, the app being temporarily disabled. The behaviour:
No catch-up. A missed fire is not replayed.
The next tick evaluates only the next CRON match — past matches are not retroactively delivered.
For an
every 5 minutesschedule that misses one tick, you simply have a 10-minute gap between deliveries instead of 5.For a
daily at 09:00 UTCschedule that misses that tick window entirely (extremely rare), the next fire is the next day at 09:00 UTC.
If the receiver depends on a guaranteed daily delivery, build idempotency on the receiver side and accept that very occasionally a day might be missed. Or pair with an event-driven webhook on a meaningful Jira event.
Minute field on a 5-minute grid
The 5-minute platform tick means the minute field of your CRON must be on the 5-minute grid:
✅
0,5,10,15, …55✅
*/5,*/10,*/15,*/30❌
1,2,3,7,13, etc.
Off-grid minutes are silently dropped — the schedule will never fire. The form's Builder validator blocks this on save, but custom-cron pasted from elsewhere can slip through. If your schedule never fires, check the minute field first.
The form preview's Cronstrue label will still say "*at HH:03*" even for an invalid minute — the label is correct about what you wrote, not about what the platform will execute.
What if your schedule never fires
Walk through this checklist:
Webhook is enabled. Look at the Status lozenge on the Webhooks list.
Schedule toggle is on in the form.
JQL is set on the Events tab.
JQL actually returns issues at the time of day the schedule runs. Run the JQL in Jira's search bar to confirm.
Minute field is on the 5-minute grid.
You waited the full 5 minutes after the planned UTC time.
The CRON expression really means what you think. Read the Cronstrue label. Convert the UTC time mentally to your timezone — does it match what you intended?
Open Webhook Logs filtered by this webhook. If JQL returns 0 issues and the empty-heartbeat toggle is off, the delivery is intentionally skipped and leaves no log row. Try toggling on Send request even if JQL returns no issues to confirm the schedule is at least firing.
Full diagnostic walkthrough: CRON didn't fire when expected (UTC pitfalls).
Related
Scheduled webhooks — the hub.
CRON syntax (Quartz, 7 fields) — the full operator reference.
Schedule examples — copy-paste with UTC↔local conversion.
CRON didn't fire when expected (UTC pitfalls) — diagnostic playbook.