Protractor 7.0.0: Handling Retries Without Duplicate Writes

Confirmed Facts

Protractor 7.0.0 does not provide native retry configuration. Retries are added via Jasmine‑level packages such as jasmine-retry or jasmine-fail-fast, which re‑execute the entire it block on failure. Because the whole block runs again, any state‑changing operation (DB insert, API call, file write) is repeated unless the test is idempotent or cleaned up.

Likely Explanation

If a test fails and is retried, the framework has no hook to snapshot or restore application state before the second run. Zone.js may hide asynchronous errors on the first attempt, making them appear only on the retry and causing non‑deterministic outcomes. Therefore, duplicate side effects arise from re‑running the same non‑idempotent code, not from a missing Protractor‑level feature.

Recommended Steps for This Case

1. Make every state‑changing operation idempotent (e.g., use upserts, idempotency keys, or conditional writes) so that repeating it does not create duplicates.
2. If idempotency is not feasible, add an afterEach hook that removes or rolls back any test data created in the it block (delete DB rows, delete temp files, cancel API resources).
3. Install a Jasmine retry package, for example:

npm install --save-dev jasmine-retry

Then enable it in protractor.conf.js:

onPrepare: function () {
  require('jasmine-retry');
  jasmine.getEnv().configure({
    retryCount: 2,
    retryDelay: 500
  });
}

4. Run the test and verify that the afterEach cleanup executes on each attempt (add a console.log or temporary marker to confirm).

Missing Diagnostic Detail

Are your database/API/file writes already idempotent, or can they be made so with minimal changes? Knowing this determines whether you need only the retry package and cleanup hooks, or whether a more complex state‑snapshot plugin would be required.