Use Retry:2 on script failure for composer jobs

Migrated issue

Reported by: cmlara

Related to !282 (merged)

Problem/Motivation

We have been experiencing random timeouts with the composer stage of gitlab_templates. The cause appears to either be external service blocks or internal infrastructure problems.

Investigation work into the cause and solution has been de-prioritized by the Drupal Infra team

Retrying failed jobs manually by human intervention is high cost.

Automatically retrying may not resolve the issues, especially if the job executes on the same runner, however it may at least increase the chance we have a successful completion

Cons:
Retry may increase the failure rates if we are triggering blocks for abusive connections.

Steps to reproduce

Proposed resolution

composer:
  retry:
    max: 2
    when:
      - script_failure
      - unknown_failure
      - api_failure
      - stuck_or_timeout_failure
      - runner_system_failure
      - scheduler_failure

Remaining tasks

User interface changes

API changes

Data model changes

Assignee Loading
Time tracking Loading