Scheduled GitHub Actions Workflow Silence and Failure Report
Pricing
from $20.00 / 1,000 run starteds
Watches scheduled GitHub Actions workflows and reports the ones that stopped to fire, that fail again and again, or that GitHub disabled for inactivity. One dataset record for each workflow plus one summary record. HTTP only, no browser, no proxy.
1 to 100 records. Each record holds owner_repo (for example "python/cpython"), workflow_file (the file name of the workflow, for example "stale.yml", or its numeric workflow id) and expected_max_interval_minutes (the longest silence that is still normal for this schedule). A workflow that did not run inside that interval is reported as overdue.
[ { "owner_repo": "python/cpython", "workflow_file": "stale.yml", "expected_max_interval_minutes": 1440 }, { "owner_repo": "home-assistant/core", "workflow_file": "stale.yml", "expected_max_interval_minutes": 1440 }, { "owner_repo": "rust-lang/rust", "workflow_file": "ghcr.yml", "expected_max_interval_minutes": 2880 }]Optional. Supply your own token to watch private repositories and to lift the rate limit of 60 requests for each hour that applies to anonymous calls. The Actor never buys a token and never sends the token anywhere except to api.github.com.
The Actor raises a failure alert after this many failed runs in a row.
Optional. One public HTTPS URL. The Actor sends one JSON POST with every alert of the run. Private, loopback and reserved addresses are refused. Leave it empty to read the alerts in the dataset only.
Which runs count as a firing of the schedule. "schedule" reads only the runs that the cron trigger started, which is the correct choice for a nightly job. "any" reads every run of the workflow.
How many recent runs are read for each workflow. The failure streak cannot become longer than this number.
Keep one cursor for each workflow in the key-value store, so the same alert is not sent again on every schedule tick. A changed alert reason or a new run always sends a new alert.
Name of the named key-value store that holds one alert cursor for each workflow. The store must be named, because the default store of an Actor belongs to one run only and would lose the cursor on every tick. Give two schedules two different names when they must alert on their own.
Read the run list a second time before a silence alert is raised. GitHub sometimes answers the run list from a replica that is weeks out of date, which reads exactly like a schedule that stopped to fire. This second read removes that false alert. It costs one extra API call, and only for a workflow that already looks silent.
How long the Actor waits before it reads the run list of the silent-looking workflows again. The GitHub run list is served with a cache lifetime of 60 seconds, so a second read without a pause can be answered from the same page as the first. The pause happens once for the whole run, and not once for each workflow.
Timeout for one call to the GitHub REST API.
Rate limit for the calls to api.github.com. The calls are sent one after the other, so this Actor cannot act as a load generator.