@lbennett/transient-tantrum

Target a GitLab CI job to retry N times to find transient failures.

This CLI tool is intended to be used for MRs that include potential fixes for transient failures on GitLab CI.

If you think you have found a fix for a transient failure, you can retry the job a number of times to increase your confidence in the fix. This does replace debugging and understanding the fix.

As of v1.1.0, you can also invert the operation and retry on failure instead of success. This was not the original intended use, but it comes in handy with known flaky tests that have no immediate fix where all you can do is retry until your pipeline is green.

Installation

npm install --global @lbennett/transient-tantrum
#OR
yarn global add @lbennett/transient-tantrum

Usage

First you must create a GitLab personal access token.

GITLAB_PRIVATE_ACCESS_TOKEN=YOUR_GITLAB_PRIVATE_ACCESS_TOKEN transient-tantrum <projectPath> <mergeRequestId> <jobName> [retryCount] [-wivh]

• projectPath: The namespace name and project name, separated by a slash. e.g. gitlab-org/gitlab-ce
• mergeRequestId: The Merge request ID. e.g. 23033
• jobName: The name of the job to test for transients. e.g. karma
• retryCount: The number of times you would like to retry the job on success. Default: 5
• waitDuration: The duration in minutes to wait between checking the job status. Default: 3
• invert: Retry on failure instead of success Default: false

$ transient-tantrum -h
transient-tantrum <projectPath> <mergeRequestId> <jobName> [retryCount]

start a transient tantrum on a specific job for a given merge request

Positionals:
  projectPath     Path to the project containing the MR to test e.g.
                  gitlab-org/gitlab-ce.
  mergeRequestId  ID of the Merge request to test e.g. 1234.
  jobName         Name of the job to retry, e.g. karma, or "docs lint".
  retryCount      Number of times to retry the job if it succeeds                   [default: 5]

Options:
  --version          Show version number                                             [boolean]
  -w, --waitDuration The duration in minutes to wait between checking the job status [number] [default: 5]
  -i, --invert       Retry on failure instead of success                             [boolean] [default: false]
  -h, --help         Show help                                                       [boolean]

Example

$ transient-tantrum gitlab-org/gitlab-ce 23033 karma 1


Starting tantrum...


Getting latest pipeline for gitlab-org/gitlab-ce!23033...
Getting pipelines for gitlab-org/gitlab-ce!23033...
Finding karma job...
Getting jobs for pipeline #36508658...
Getting jobs for pipeline #36508658...
Getting jobs for pipeline #36508658...
The karma jobs status is currently "running".
Checking pipeline in 3 minutes...
Getting job...
The karma jobs status is currently "running".
Checking pipeline in 3 minutes...
Getting job...
The karma jobs status is currently "running".
Checking pipeline in 3 minutes...
Getting job...
The karma jobs status is currently "running".
Checking pipeline in 3 minutes...
Getting job...
The karma jobs status is currently "success".
Retrying job...
Checking pipeline in 3 minutes...
Getting job...
The karma jobs status is currently "running".
Checking pipeline in 3 minutes...
Getting job...
The karma jobs status is currently "running".
Checking pipeline in 3 minutes...
Getting job...
The karma jobs status is currently "running".
Checking pipeline in 3 minutes...
Getting job...
The karma jobs status is currently "success".
Tantrum complete.
✨  Done in 1465.05s.