Cron tools › Explain / Test
Cron DST checker
Pick the scheduler and the time zone it runs in. The checker finds every daylight-saving change in the next twelve months and lists the runs that land in the skipped or repeated hour, with what that scheduler does to each one — or "not documented" where the scheduler does not say.
Rule used
Clock changes
Try one
How to check a schedule for daylight-saving problems
- Type the cron expression in the syntax of your scheduler: five fields for Linux cron,
cron( … )with six fields for EventBridge, seconds first for node-cron if you use them. - Pick the Scheduler. Its rule for skipped and repeated hours is shown under Rule used, with where it comes from.
- Pick the Time zone the scheduler evaluates the expression in: the server's zone for Linux cron, the
timezoneoption or field for the others. For an always-UTC platform, pick your own zone instead. - Read each clock change. A run marked skipped does not happen that day, runs twice happens in both passes of the repeated hour, moved runs at the moment the clocks jump, and not documented means the scheduler does not say.
- Change Starting to look further ahead, or Copy link to share the check.
What schedulers do when the clocks change
| Scheduler | A time the clocks skip | A time that happens twice |
|---|---|---|
| Linux cron (cronie) | Fixed-time jobs run right after the jump; jobs with * at the start of the minute or hour field do not run | Fixed-time jobs run once (first pass); wildcard jobs run in both passes |
| node-cron 4.x | Skipped | Runs on the first occurrence only (documented) |
GitHub Actions with timezone: | Advances to the next valid time (2:30 AM → 3:00 AM) | Not documented |
| EventBridge Scheduler | Skipped | Runs once, at the second occurrence, after the shift |
| Google Cloud Scheduler | Google warns that DST "can cause jobs to run or not run unexpectedly" and recommends UTC | |
| Kubernetes, GitLab, Vercel Schedules | Not documented | |
UTC only: Vercel Cron Jobs, Cloudflare Workers, EventBridge rules, GitHub without timezone: | Nothing is skipped or repeated; the local time of every run moves by the size of the change | |
One documented rule does not always match the code. node-cron's documentation says a repeated hour runs on its first occurrence, but node-cron 4.6.0 fired the second occurrence when we ran it in zones ahead of UTC (London, Berlin, Kyiv, Sydney, Lord Howe, Auckland) and the first in zones behind it (New York, Chicago, Los Angeles, Santiago). The checker shows the documented rule and flags the rows where the observed behaviour differs.
Why Linux cron treats 30 2 * * * and */15 * * * * differently
cronie decides by the first character of the minute and hour fields. If either starts with * the job is a "wildcard" job,
otherwise it is a "fixed-time" job (@hourly counts as wildcard, the other @ shorthands as fixed-time). When the
clock jumps forward, cron runs the fixed-time jobs of every skipped minute straight away and lets the wildcard jobs carry on from the new
time. When the clock goes back, it runs the wildcard jobs again and holds the fixed-time jobs back, because they "probably have already
run". So in New York a 02:30 backup runs at 03:00 on the spring-forward night, an every-15-minutes job simply has no 02:00–02:45 runs,
and on the fall-back night the 15-minute job runs eight times between 01:00 and 02:00 while a 01:30 job runs once.
Two details catch people out. 0 */2 * * * starts its hour field with *, so it is a wildcard job even though it
runs only every two hours. And cronie's own source warns that jobs using CRON_TZ in a zone other than the system's can be
"missed or run twice during the DST change in the job timezone". Changes of more than three hours are treated as clock corrections,
not DST.
Zones without daylight saving time
Asia/Tokyo, Asia/Kolkata and UTC have no clock changes, so the checker finds nothing to report: the safest fix for a schedule that must not skip or double is to run the scheduler in UTC, or in a zone without DST. In the southern hemisphere the order is reversed — Australia/Sydney goes back an hour in April and forward in October — and a few zones change by 30 minutes (Australia/Lord_Howe) or at midnight (America/Havana). The transitions come from your browser's copy of the IANA time zone database.
For the full list of upcoming runs around a change, use the next run calculator; to move a schedule to another zone, the time zone converter; and for a whole crontab, the crontab checker.
FAQ
Will my 02:30 cron job run on the night the clocks go forward?
On Linux cron (cronie, Debian) yes: a fixed-time job whose time was skipped runs right after the jump, at 03:00. node-cron and
EventBridge Scheduler skip it that day; GitHub Actions with timezone: moves it to 03:00. Others do not document it.
Will a 01:30 job run twice when the clocks go back?
Not on Linux cron: "running the same job twice is avoided" for fixed-time jobs, so it runs in the first pass only. A job with
* in its minute or hour field does run in both passes. EventBridge Scheduler runs it once, at the second 01:30.
Why does the checker say "not documented"?
Because the scheduler's own documentation does not say what happens, and the checker only shows behaviour a primary source states. For those schedulers the rows tell you which runs are at risk, so you can move them out of the changing hour.
My platform only runs in UTC. Why does it still show clock changes?
A UTC schedule never skips or repeats, but your local time moves around it. A job at 14:30 UTC is 09:30 in New York in winter and 10:30 in summer; the checker shows the last run before and the first run after each change in your zone.