Skip to content

chore(system-tests): retry Docker image pulls on transient registry errors#34147

Open
cursor[bot] wants to merge 1 commit into
developfrom
cursor/pipeline-status-and-credits-5838
Open

chore(system-tests): retry Docker image pulls on transient registry errors#34147
cursor[bot] wants to merge 1 commit into
developfrom
cursor/pipeline-status-and-credits-5838

Conversation

@cursor

@cursor cursor Bot commented Jun 22, 2026

Copy link
Copy Markdown
Contributor

Summary

Develop CI for commit 19ac719 failed binary-system-tests when Docker Hub returned HTTP 500 while pulling cypress/base:25.0.0. This blocked ready-to-release even though the merged PR (#34141) only removed a duplicate webpack-dev-server test invocation.

Add retry with exponential backoff to DockerProcess.pull() so transient Docker registry failures do not fail the binary system test suite.

  • Closes N/A (CI flake observed on develop pipeline 83085)

Additional details

Failure excerpt from binary-system-tests job #3735965:

Error: (HTTP code 500) server error - Head "https://registry-1.docker.io/v2/cypress/base/manifests/25.0.0": unknown

prePullImages already uses allSettled, but individual test runs call pull() again via dockerSpawner. Retries are applied at the pull level (3 attempts, 2s/4s backoff).

Unrelated failures on the same pipeline (Windows port 4455 conflict and studio UI flake) are not addressed here.

Steps to test

  • [na] — harness-only change; validated by existing binary-system-tests CI job on next develop run.

How has the user experience changed?

No change

PR Tasks

  • Is there an associated issue with maintainer approval for PR submission?
  • Have tests been added/updated?
  • Has a PR for user-facing changes been opened in cypress-documentation?
  • Have API changes been updated in the type definitions?
Open in Web View Automation 

Note

Low Risk
CI harness-only change; no production runtime, auth, or user-facing behavior.

Overview
Docker image pulls in system tests now retry instead of failing on the first registry error (e.g. Docker Hub HTTP 500).

The previous single-shot pull logic is moved to pullOnce(). pull() runs up to 3 attempts by default, logging failures and waiting 2s then 4s between tries before rethrowing the last error. Call sites such as dockerSpawner and prePullImages keep calling pull(), so binary system tests get retries without other harness changes.

Reviewed by Cursor Bugbot for commit 1f044fd. Bugbot is set up for automated code reviews on this repo. Configure here.

binary-system-tests failed on develop (19ac719) when Docker Hub returned
HTTP 500 for cypress/base:25.0.0. Add exponential backoff retries to
DockerProcess.pull() so transient registry failures do not fail the suite.

Co-authored-by: Cacie Prins <cacieprins@users.noreply.github.com>
@cypress-app-bot

Copy link
Copy Markdown
Collaborator

@jennifer-shehane jennifer-shehane marked this pull request as ready for review June 22, 2026 14:11
@jennifer-shehane jennifer-shehane changed the title fix(system-tests): retry Docker image pulls on transient registry errors chore(system-tests): retry Docker image pulls on transient registry errors Jun 22, 2026
@jennifer-shehane jennifer-shehane requested a review from a team June 22, 2026 16:46
return
} catch (err) {
lastError = err
if (attempt < maxAttempts) {

Copy link
Copy Markdown
Collaborator

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

This makes sense to me. Do we have a way to unit test this and verify it works?

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

3 participants