Guides
When checks fail
Timing
Due checks are picked up every 15 seconds. Each check also runs up to 10% of its interval early or late, at most 5 minutes, so watchers made together don't all fetch the site at once. An hourly check runs every 55 to 65 minutes, and a 15-minute one every 13.5 to 16.5 minutes.
List runs shows when each check was due (due_at) and how many seconds late it ran (late_seconds). Older checks have null for both.
Blocks
A check is blocked when the site refuses it (a 403, a 429 or an unsolved captcha), unreachable when the address doesn't answer, and not_found when the page is gone.
- Nothing fires from a site that couldn't be read, because an unknown condition never passes.
- The site's last good results are kept, so a restock is still compared with real data once it answers again.
- After 3 in a row, you get a
blockedevent with the outcome asreasonand the error asdetail. - When the site can be read again, you get
recoveredwithwas: "blocked". - These checks cost no credits, and checks keep running on schedule.
Failed checks
A check is failed when the site couldn't be read for another reason: the page or API changed shape, the code that reads it broke, robots.txt refuses it, or the daily cap ran out. These cost no credits either.
After 3 in a row, you get a failing event with the cause as reason and the error as detail. These reasons send it on the first check, because waiting won't fix them:
reason | Means | What to do |
|---|---|---|
robots_disallowed | The site's robots.txt disallows the page | Watch a different page, or set fetch.accept_site_terms if the site's terms let you fetch it. See Sites |
usage_limit | Today's browser runs with steps (200) or captcha solves (100) are used up | Wait for the reset at 00:00 UTC, or check less often |
Other common reasons:
reason | Means |
|---|---|
unexpected_answer | A site with built-in support changed the shape of its page or API. The check fails so the page isn't taken as empty |
recipe_broken | A page read by a learned recipe stopped finding its items. After 3 in a row, the recipe is learned again, at most once every 6 hours |
step_failed | A browser step couldn't be done. detail says which one |
private_address | The page's address is on a private or local network |
When the site can be read again, you get recovered with was: "failing". Blocked and failed checks each count their own 3 in a row.
Every issue code is in check issues.
When a watcher ends
A watcher stops and sends ended once when it reaches its until, or after 24 hours on a test key. With until: "after_first_match", it sends matched when it stops and no ended.
False matches
A price that's wrong for one check can set off a match. Set notify.confirm to the number of checks in a row a match must hold for:
"notify": { "channels": ["webhook"], "confirm": 2, "webhook": { "url": "https://you.example/snag", "events": ["matched"] } }confirm is 2 to 5. Using it with conditions that are only true on one check (added, removed, changed, judge and back_in_stock) returns 422 confirm_not_supported.
While a match waits, that check's issues include awaiting_confirmation.
See every check
List runs shows every check, what it read and why it fired or didn't. The dashboard shows the same under Checks on each watcher. Each site's outcome is one of these:
outcome | Means |
|---|---|
ok | The site was read. Only these checks cost credits |
blocked | The site refused the check with a bot wall, an unsolved captcha or a rate limit |
unreachable | The site didn't answer, timed out or answered with an error |
not_found | The page is gone (a 404 or 410) |
failed | The site couldn't be read for another reason, like a changed page, robots.txt or the daily cap. issues says why |