Skip to main content

Run Playwright tests on a schedule

Scheduled runs are where CI pipelines get duplicated. The nightly regression is usually a copy of the pull-request workflow with a different trigger and a few things commented out, and the two drift apart within a month.

A Schedule node makes the trigger part of the graph instead of part of the pipeline file, so the same workflow can be started by a schedule and by anything else.

What you need​

Build the workflow​

Schedule ──sequential──▶ Environment ──sequential──▶ Playwright

The Schedule node starts the run on a recurring cadence. Everything downstream is the workflow you would have built anyway.

Note that Schedule is a trigger node and does not accept inbound connections — nothing can point at it. The connection state rules explain how the editor derives this from node metadata; in practice the editor simply will not offer it as a connection target.

Choosing a cadence​

  • Nightly suits a full regression suite that takes too long for every pull request.
  • Hourly or more often suits a small smoke suite against a live environment, where the point is detecting an outage rather than a regression.
  • Weekly suits long-running or expensive suites — cross-browser matrices, or anything hitting a rate-limited third party.

Set the cadence so a run has comfortably finished before the next one starts. Each node runs at most once per execution, but two overlapping executions of the same workflow against the same environment will interfere with each other at the application level.

Running against different environments​

Rather than duplicating the workflow per environment, branch after the schedule:

                ┌──concurrent──▶ Environment (staging)  ──▶ Playwright
Schedule ───────┤
                └──concurrent──▶ Environment (preprod)  ──▶ Playwright

Concurrent edges start both branches at once, so the two environments are tested in parallel rather than one after the other. Use Sequential instead if they share a resource that cannot handle both.

Alerting on a scheduled run​

Nobody watches a nightly run. Attach the notification to it directly:

Schedule ──▶ Environment ──▶ Playwright ──failure──▶ Slack

See Slack alerts for failed tests. For a nightly suite specifically, routing the failure through AI triage first is worth it — someone reads that message cold the next morning with no context.

Versus on: schedule in CI​

A cron entry in CI works. It stops being pleasant when the scheduled variant needs different environment variables, a different browser matrix, different retry behaviour, and a notification the pull-request version does not have — at which point it is a second pipeline to keep in sync. Here the schedule is one node in front of a workflow that already exists.

Next steps​