These shared workflows implement the commmon patterns of how we deploy web applications:
- Lint Python and JavaScript Code
- Test JavaScript Code
- Build a Test Image using the app's Dockerfile so that tests run in a near production like environment
- Upload coverage information from JS Tests and Python tests to Codecov
- Build a Production image and push it to
ghcr.io, taggedsha-<commit-sha> - if on
main, Deploy to QA - if a Release, deploy that same
sha-<commit-sha>image to Production — no rebuild. A release of a commit that never landed onmainfails rather than building one. - Wait for AppPack's Build, Test, Finalize, Release, and Deploy phases before marking the deployment successful.
- assumes use of Namespace.so runners which are far better than GitHub's included ones (maybe there is a way to make this configurable so others who want to use stock runners can do so -- PRs welcome!)
Here are examples for an AppPack project with QA/prod apps called my-apppack-qa-app and my-apppack-prod-app.
CODECOV_TOKENis your token from Codecov.CR_UNandCR_PATare your username and personal access token for pushing to GitHub Container Registry.SENTRY_AUTH_TOKENis your auth token from Sentry withproject:writepermissions for uploading source maps.APPPACK_ROLE_ARNis the AWS role assumed through GitHub OIDC for deployments.
These need to be set as secrets in your GitHub repository.
Pass more BuildKit secrets to test.yml with BUILD_SECRETS, one NAME=value per line, built from your own repo secrets:
secrets:
BUILD_SECRETS: |
MAXMIND_LICENSE_KEY=${{ secrets.MAXMIND_LICENSE_KEY }}Each entry is mounted in both the test and production builds as id=NAME:
RUN --mount=type=secret,id=MAXMIND_LICENSE_KEY cat /run/secrets/MAXMIND_LICENSE_KEYNames must match ^[A-Z][A-Z0-9_]*$. Values are masked in logs, never become build args or image layers, and cannot contain newlines. Empty values (a missing repo secret, Dependabot runs) are skipped with a warning.
# myapp/.github/workflows/test.yml
name: Test and Build
on:
push:
branches: "**"
tags-ignore: "**"
jobs:
test-and-build:
name: CI
uses: wedgworth/actions/.github/workflows/test.yml@main
with:
python-src-dir: myapp
secrets:
CODECOV_TOKEN: ${{ secrets.CODECOV_TOKEN }}
CR_UN: ${{ secrets.CR_UN }}
CR_PAT: ${{ secrets.CR_PAT }}
SENTRY_AUTH_TOKEN: ${{ secrets.SENTRY_AUTH_TOKEN }}
deploy-qa:
name: CD
needs: [test-and-build]
if: ${{ github.event.ref == 'refs/heads/main' }}
uses: wedgworth/actions/.github/workflows/deploy.yml@v13.0.0
with:
app-name: my-apppack-qa-app
image: ghcr.io/${{ github.repository }}:${{ needs.test-and-build.outputs.version }}
secrets:
APPPACK_ROLE_ARN: ${{ secrets.APPPACK_ROLE_ARN }}
CR_UN: ${{ secrets.CR_UN }}
CR_PAT: ${{ secrets.CR_PAT }}# myapp/.github/workflows/release.yml
name: Release
on:
release:
types: [published]
jobs:
release:
uses: wedgworth/actions/.github/workflows/release.yml@v13.0.0
with:
app-name: my-apppack-prod-app
secrets:
CR_UN: ${{ secrets.CR_UN }}
CR_PAT: ${{ secrets.CR_PAT }}
APPPACK_ROLE_ARN: ${{ secrets.APPPACK_ROLE_ARN }}deploy.yml owns image setup, AWS authentication, deployment, and all five
tracking steps. QA calls it directly. release.yml is a small wrapper that calls
the same workflow with promote-release: true and the tested sha-<commit> image.
Promotion stamps the GitHub release tag into the image and retrieves the original
CI logs before continuing through the common deployment steps. The wrapper's
relative workflow reference uses the same commit/tag as the wrapper itself.
Before upgrading deployment workflows from v12, add these read permissions to
the existing deployment role. Replace <AWS_REGION>, <AWS_ACCOUNT_ID>,
<QA_APP_NAME>, and <PROD_APP_NAME> with your values. Keep the literal APP#
prefix in the DynamoDB keys:
{
"Version": "2012-10-17",
"Statement": [
{
"Effect": "Allow",
"Action": "dynamodb:GetItem",
"Resource": "arn:aws:dynamodb:<AWS_REGION>:<AWS_ACCOUNT_ID>:table/apppack",
"Condition": {
"ForAllValues:StringEquals": {
"dynamodb:LeadingKeys": ["APP#<QA_APP_NAME>", "APP#<PROD_APP_NAME>"]
}
}
},
{
"Effect": "Allow",
"Action": "codebuild:BatchGetBuilds",
"Resource": [
"arn:aws:codebuild:<AWS_REGION>:<AWS_ACCOUNT_ID>:project/<QA_APP_NAME>",
"arn:aws:codebuild:<AWS_REGION>:<AWS_ACCOUNT_ID>:project/<PROD_APP_NAME>"
]
}
]
}The workflows verify these permissions before starting a deployment. Tracking uses
the exact build_number and build_arn returned by apppackio/deploy-action,
reading the same DynamoDB record as the AppPack CLI.
CodeBuild success only completes the build portion; the Deploy step requires
AppPack's deploy.state to be succeeded. Failed phases (including a configured
postdeploy command), stopped CodeBuild jobs, and timeouts fail CI. Failure
annotations include AppPack's log location when available. Build/test logs remain
in the existing CI steps and AppPack; the tracker reports statuses and durations.
Each phase polls every 10 seconds with a 20-minute limit; the job has a 45-minute limit. Test and Release can be reported as skipped in the step log/summary if AppPack advances without running them. Deploy always requires explicit success. Jobs using these workflows serialize deployments per repository and app. External deployments and workflow cancellation are outside this concurrency group; cancelling/timing out CI stops monitoring, not the running AWS deployment.
The tracker uses Python 3 and the AWS CLI. The preflight action installs AWS CLI
v2 if it is absent on a Linux runner. No AppPack dashboard login is required.
Run its regression tests with python3 -m unittest discover -s tests -v.
When publishing v13.0.0, include both deployment workflows and the
.github/actions/apppack-status directory in that tag: workflows pin the action
to the same release version.