Quarantine and re-run

Two tools for a noisy suite:

• Quarantine keeps an unreliable test running and reporting, but stops it from failing CI or sending alerts. Use it while you fix a flaky test, or while a dependency it needs is down.
• Rerun failed queues only the tests whose latest run failed, instead of the whole project. Use it after a fix, or after an outage that broke many tests at once.

Quarantine a test

A quarantined test behaves like this:

	Quarantined test
Runs on schedule, from CI and with Run all	Yes
Shows its result, steps and evidence	Yes
Can fail the CI gate	No. Its failures are counted separately as quarantined_failed.
Sends Slack, webhook or email alerts	No
Retried automatically	No
Appears in Triage	Yes, with a Quarantined badge

From the app

Any of these work. All need the editor role.

• Triage → Flaky: click Quarantine on the test’s row.
• The test’s Settings tab: Reliability → Quarantine, then Save changes.
• The project board: open the test’s quick view and turn on Quarantine.

To release a test, use Release on Triage → Quarantined, or turn the switch off again.

The Quarantined tab lists who quarantined each test, when, and the note they left, plus the test’s latest result.

From the API

POST /api/reliability/test-cases/{test_case_id}/quarantine takes quarantined and an optional note (up to 500 characters) and returns the updated test.

bash

curl -fsS -X POST "$PLAINWRIGHT_URL/api/reliability/test-cases/$TEST_ID/quarantine" \
  -H "Authorization: Bearer $PLAINWRIGHT_TOKEN" -H "Content-Type: application/json" \
  -d '{"quarantined": true, "note": "Payment sandbox is down - see SHOP-142"}'

PowerShell

Invoke-RestMethod -Method Post `
  -Uri "$env:PLAINWRIGHT_URL/api/reliability/test-cases/$env:TEST_ID/quarantine" `
  -Headers @{ Authorization = "Bearer $env:PLAINWRIGHT_TOKEN" } -ContentType 'application/json' `
  -Body '{"quarantined": true, "note": "Payment sandbox is down - see SHOP-142"}'

Plainwright records who quarantined the test and when in the test’s config.quarantine ({at, by_user_id, by_name, note}). Releasing the test removes that record.

Quarantine is a stopgap

A quarantined test still runs, so it still costs a run every time. Its failures no longer reach anyone, so a real regression can hide there. Leave a note that says why, and review the Quarantined tab regularly.

Quarantine or pause?

	Quarantine	Pause
Where	Test → Settings → Reliability	Test → Settings (pause switch)
Runs	Yes	No. Paused tests are skipped by Run all, CI, schedules and Rerun failed.
Can fail CI or alert	No	It doesn’t run, so no
Use it when	You still want to see results while you fix it	The test should not run at all for now

Re-run only the failures

Rerun failed finds every active test whose newest run ended as failed or error and queues one new run for each, all in a new batch. A human verdict override counts: a run marked as passed is not re-run. Tests whose newest run is still pending or running are skipped, because they are already running.

1. Pick the environment in the top bar if you don’t want the project’s default URL. Rerun failed uses it.
2. Click Rerun failed on the Triage page, or on the project board toolbar.
3. Plainwright tells you how many runs it queued. Follow them on the Runs tab.

From the API or CI

POST /api/reliability/project/{project_id}/rerun-failed needs the editor role. Every query parameter is optional:

Parameter	Default	Meaning
environment_id	project base URL	Run against this environment. It must belong to the project (404 if it doesn’t exist, 400 if it is another project’s).
browser	chromium	chromium, firefox or webkit.
device	desktop	desktop, iphone, pixel or ipad.
batch_id	none	Only consider runs from this batch, taking the newest attempt per test. Use it to re-run the failures of one CI run.

curl -fsS -X POST \
  "$PLAINWRIGHT_URL/api/reliability/project/$PROJECT_ID/rerun-failed?batch_id=$BATCH_ID" \
  -H "Authorization: Bearer $PLAINWRIGHT_TOKEN"

Response

{
  "batch_id": "5f0c1d2e3a4b4c5d9e8f7a6b5c4d3e2f",
  "count": 3,
  "run_ids": ["…", "…", "…"],
  "status_url": "/api/batches/5f0c1d2e3a4b4c5d9e8f7a6b5c4d3e2f"
}

When nothing has failed, the response is {"batch_id": null, "count": 0, "run_ids": [], "status_url": null}. Poll status_url exactly like a CI batch to get a pass/fail gate; see CI webhook and gates.

Note

This endpoint needs a user token (Authorization: Bearer …). The project webhook secret only works for the CI webhook and for reading batch status.

Next steps

Retries and flaky testsRetry failures automatically and see which tests are flaky.

CI webhook and gatesHow quarantined tests and retries affect the deploy gate.

Triage and issuesFind out what the failures have in common before you re-run them.