Go tests for standalone Steve: CRUD, filtering, sorting, pagination, and SQL cache behavior.
Requires Go 1.27.2+.
go test -p=4 -parallel=4 ./...Live tests skip unless a Steve endpoint is configured. Copy steve-config.yaml.example to ~/steve-config.yaml, set the endpoint, and run the full suite:
STEVE_CONFIG="$HOME/steve-config.yaml" \
STEVE_TEST_ALLOW_MUTATIONS=true \
STEVE_TEST_ALLOW_WORKLOADS=true \
STEVE_TEST_ALLOW_CRDS=true \
go test -count=1 -p=4 -parallel=4 -timeout=20m -v ./...Omit the permission flags for read-only checks. Tests clean up their resources. Set STEVE_TOKEN if authentication is required; use STEVE_CA_FILE for a private CA or STEVE_INSECURE_SKIP_VERIFY=true for a local self-signed endpoint. STEVE_URL can override the configured endpoint.
Set STEVE_VERSION for local runs to use version-specific expectations and skip unsupported features. GitHub Actions sets it automatically.
In GoLand, select All Steve tests and click Play. It uses ~/steve-config.yaml, enables all test permissions, and shows individual results. The TestAll wrapper also runs the suite but appears as one test.
Requires Docker, k3d, kubectl, jq, and Go; gotestsum is optional. This command builds the selected Steve tag, creates a disposable k3d cluster, runs the full suite, and tears it down:
STEVE_VERSION=v0.10.9 scripts/e2e.sh allTo keep the environment between runs:
STEVE_VERSION=v0.10.9 scripts/e2e.sh up
scripts/e2e.sh test
E2E_RUN='TestPods' scripts/e2e.sh test
scripts/e2e.sh downLogs and test results survive teardown. The runner enables and verifies SQL cache mode. Run scripts/e2e.sh without arguments for all settings.
| Workflow | Purpose |
|---|---|
| Steve E2E Tests | Manually test a Steve tag on k3d, save results and diagnostics, and optionally report to Qase. |
| Upload Steve Qase Schemas | Create/update Qase suites and cases before the first reported run and after schema changes. |
| Check Steve Tags | Check for new tags daily at 14:00 UTC or manually; dispatch E2E runs with Qase enabled. |
| Steve Tests CI | Check workflows, shell scripts, unit tests, and schema coverage on pull requests or manually. |
For a first run, open Actions → Steve E2E Tests → Run workflow, select a branch, and enter a tag such as v0.10.9. Leave optional inputs blank for the full suite. Turn off Report results to Qase to run without secrets or AWS setup.
Manual workflows must exist on the default branch before their Run workflow button is available. No existing Kubernetes cluster is needed.
Add repository secrets under Settings → Secrets and variables → Actions:
| Secret | Value | Used by |
|---|---|---|
QASE_AUTOMATION_TOKEN |
API token with access to suites, cases, milestones, runs, and results | Qase upload/reporting |
QASE_PROJECT_ID |
Project code: RM |
Qase upload/reporting |
AWS_ROLE_ARN |
IAM role trusted by GitHub OIDC | Tag watcher |
AWS_REGION |
Cache bucket's region | Tag watcher |
STEVE_TAGS_S3_BUCKET |
Bucket name, without s3:// |
Tag watcher |
After adding the Qase secrets, run Upload Steve Qase Schemas, then Steve E2E Tests with reporting enabled. Leave the run ID blank to create a new run. Runs close only when tests and reporting both succeed.
Each top-level test under tests/ maps to one Qase case in steve_schemas.yaml, under Go Automation/Steve/<package>. Subtest counts and failures appear in the parent's result comment; internal harness tests are excluded. Qase totals therefore differ from the Go test total.
The schemas target RM; changing projects requires updating their projects entries as well as the secret. RM runs require custom field 16 with value 1 (RecurringRun). Milestones use the prefix Steve.
Create a private S3 bucket with S3-managed encryption and an IAM role using GitHub OIDC. Restrict trust to the intended repository and branch, with audience sts.amazonaws.com; match the repository's immutable subject format if enabled. Grant:
s3:ListBucketon the bucket.s3:GetObjectands3:PutObjectonsteve/*in that bucket.
Add the three AWS secrets and run Check Steve Tags. Repository policies must allow id-token: write and actions: write. Static AWS keys and a GitHub PAT are unnecessary.
- Tracks the highest Steve tag, including prereleases, for each minor used by the five newest Rancher release branches plus
main. Set theRANCHER_RELEASE_LINESrepository variable to change the window (default5). - A minor without a cached tag dispatches on the next scan; later scans dispatch only changed tags. Shared Steve minors produce one dispatch. Older tags are not backfilled.
- Always dispatches E2E runs from the repository's default branch.
- Stores dispatched tags in
steve/<minor>.txt; test artifacts stay in GitHub. Keep cache objects out of expiration policies. Failed tests need a manual retry; manual E2E runs do not update the cache. - Enable scheduled workflows in forks. If a fork and upstream use the same bucket and prefix, schedule discovery in only one.
- Indexed field: add a row to TestSQLIndexedFields with fixtures and an update.
- Filter, sort, or pagination: add a resource collection scenario; see ConfigMaps.
- Different setup: add a test in the matching resource package using helpers in
actions/. - New top-level test: add its case to steve_schemas.yaml and upload the schemas.
Run one case with -run:
go test -count=1 -v ./tests/indexedfields -run 'TestSQLIndexedFields/hpa_min_replicas'Keep cases independent, use explicit expected results, and clean up resources. Use t.Parallel() only for cases that own their fixtures.