prow-trigger-nightly - SKILL.md Agent Skill

name: prow-trigger-nightly description: >- Trigger RHDH nightly ProwJobs on demand via the OpenShift CI Gangway REST API. Supports both rhdh and rhdh-plugin-export-overlays repos. Use when the user wants to trigger, run, kick off, or start a nightly CI job, run an on-demand E2E nightly test, list available nightly jobs, or trigger an overlay nightly. Also use when the user mentions Gangway, "nightly job", "periodic-ci", RC verification, testing a custom image, running CI against a fork, or checking available image tags on quay.io.

Trigger Nightly ProwJobs

Trigger RHDH nightly ProwJobs via the OpenShift CI Gangway REST API.

Supports two repositories:

rhdh — the main RHDH application (periodic-ci-redhat-developer-rhdh-*-nightly)
rhdh-plugin-export-overlays — plugin export overlays (periodic-ci-redhat-developer-rhdh-plugin-export-overlays-*-nightly)

Script Location

All commands below use paths relative to this skill's directory: skills/prow-trigger-nightly/scripts/trigger_nightly_job.py

Prerequisites

Python 3.9+
oc CLI installed (for authentication to OpenShift CI)

Flow

Fetch available jobs and let the user pick one
Ask about image override and additional options (fork, alerts)
Show the command, confirm, execute, report results

Step 1: Fetch Jobs and Select

List configured nightly jobs:

uv run scripts/trigger_nightly_job.py --list

Present the jobs in a table with columns: short name and which branches have it. Derive the short name from the job name part after the branch segment (e.g. e2e-ocp-helm-nightly -> "OCP Helm"):

Repo	Job	main	release-1.9	release-1.8
rhdh	OCP Helm	x	x	x
rhdh	AKS Helm	x	x
overlays	OCP Helm	x

Then ask the user to describe which job and branch they want in natural language.

Natural Language Mapping

Map the user's description to the matching full job name from the fetched list. If no branch is mentioned, default to main:

RHDH repo jobs:

"ocp helm" / "openshift helm" -> e2e-ocp-helm-nightly (not upgrade, not versioned)
"operator" / "ocp operator" -> e2e-ocp-operator-nightly (not auth-providers)
"helm upgrade" / "upgrade test" -> e2e-ocp-helm-upgrade-nightly
"auth providers" / "authentication" -> e2e-ocp-operator-auth-providers-nightly
"4.17", "4.19", "4.20", "4.21" -> e2e-ocp-v4-{VERSION}-helm-nightly
"aks helm" / "azure helm" -> e2e-aks-helm-nightly
"aks operator" / "azure operator" -> e2e-aks-operator-nightly
"eks helm" / "aws helm" -> e2e-eks-helm-nightly
"eks operator" / "aws operator" -> e2e-eks-operator-nightly
"gke helm" / "google helm" -> e2e-gke-helm-nightly
"gke operator" / "google operator" -> e2e-gke-operator-nightly
"osd" / "osd gcp" -> e2e-osd-gcp-helm-nightly or e2e-osd-gcp-operator-nightly
Branch: "1.9", "release 1.9", "1.8 branch" -> match from that branch
Multiple: "all AKS jobs", "all Operator jobs on main" -> offer to trigger them in sequence

Overlay repo jobs:

"overlay nightly" / "overlay helm" / "overlays nightly" -> periodic-ci-redhat-developer-rhdh-plugin-export-overlays-main-e2e-ocp-helm-nightly

Shared Cluster Constraint (GKE / OSD-GCP only)

GKE and OSD-GCP each share a single cluster — never run two jobs on the same platform simultaneously. Before triggering, warn the user.

Step 2: Options

Overlay repo jobs

Overlay jobs support fork overrides (--org, --repo, --branch), catalog index override (--catalog-index-image), and Playwright version override (--playwright-version).

Image overrides (--image-registry, --image-repo, --tag), --chart-version, and --send-alerts are NOT supported — the script will error if these are passed for an overlay job.

If the user doesn't need any overrides, skip this step and go directly to Step 3.

RHDH repo jobs

Present all options together. The user picks by number — multiple selections allowed (e.g. "2, 5"):

Image override:

Default image — no image flags, use whatever the job is configured with
Custom tag only — override just the tag, keep default registry and repo
Custom repo + tag — override image repository and tag, keep default registry (quay.io)
Fully custom image — override registry, repo, and tag

Catalog & chart override: 5. Catalog index image — override the plugin catalog index image (--catalog-index-image) 6. Chart version — override the Helm chart version (--chart-version)

Additional options: 7. Fork override — run against a fork instead of redhat-developer/rhdh 8. Send Slack alerts — notify via --send-alerts

Constraint: --image-repo requires --tag, but --tag works on its own. --playwright-version is overlay-only and will error for RHDH jobs.

Follow-up based on selections

If 2 or 3 selected (quay.io registry) — fetch available tags and present as numbered options. For release-* branches, derive --tag-filter by stripping the release- prefix. For main, omit --tag-filter to show all available versions:

# For release-1.10 branch:
uv run scripts/trigger_nightly_job.py --list-tags --tag-filter 1.10

# For main branch (show all versions):
uv run scripts/trigger_nightly_job.py --list-tags

Use --image-repo <REPO> to query a different image repository (default: rhdh/rhdh-hub-rhel9). Present the numbered results with a final option to enter a custom tag (e.g. next, latest). For option 3, also ask for the image repository.

If 4 selected (non-quay registry) — ask for all three values (tag fetching not available):

Registry (e.g. brew.registry.redhat.io)
Image repo (e.g. rhdh/rhdh-hub-rhel9)
Tag (e.g. 1.9)

If 5 selected — ask for catalog index image (e.g. quay.io/rhdh/plugin-catalog-index:1.9-60 for RC, registry.access.redhat.com/rhdh/plugin-catalog-index:1.9.4 for GA).

If 6 selected — ask for chart version (e.g. 1.9-227-CI).

If 7 selected — ask for:

GitHub org (--org): e.g. my-github-user
Repo name (--repo): e.g. rhdh
Branch (--branch): e.g. my-feature-branch

Step 3: Confirm and Execute

Show the full command and present final options:

uv run scripts/trigger_nightly_job.py \
  --job <FULL_JOB_NAME> \
  [--image-registry <REGISTRY>] \
  [--image-repo <REPO>] \
  [--tag <TAG>] \
  [--catalog-index-image <IMAGE>] \
  [--chart-version <VERSION>] \
  [--playwright-version <VERSION>] \
  [--org <ORG>] \
  [--repo <REPO>] \
  [--branch <BRANCH>] \
  [--send-alerts] \
  [--dry-run]

Execute — run the command as shown
Change something — go back and modify parameters

After execution, show the API response. If a job URL or ID is returned, display it prominently. On error, help diagnose (common issues: expired token, invalid job name).

RC Verification Example

uv run scripts/trigger_nightly_job.py \
  --job periodic-ci-redhat-developer-rhdh-main-e2e-ocp-helm-nightly \
  --image-repo rhdh/rhdh-hub-rhel9 --tag 1.9-227 \
  --catalog-index-image quay.io/rhdh/plugin-catalog-index:1.9 \
  --chart-version 1.9-227-CI

GA Verification Example

uv run scripts/trigger_nightly_job.py \
  --job periodic-ci-redhat-developer-rhdh-main-e2e-ocp-helm-nightly \
  --image-registry registry.redhat.io --image-repo rhdh/rhdh-hub-rhel9 --tag 1.9.4 \
  --catalog-index-image registry.access.redhat.com/rhdh/plugin-catalog-index:1.9.4

Reference

Script flags: -j/--job, -l/--list, -T/--list-tags, --tag-filter, -I/--image-registry, -q/--image-repo, -t/--tag, --catalog-index-image, --chart-version, --playwright-version, -o/--org, -r/--repo, -b/--branch, -S/--send-alerts, -n/--dry-run, --json
Dedicated kubeconfig at ~/.config/openshift-ci/kubeconfig — won't interfere with your current cluster context
If auth is needed, the script opens a browser for SSO login
RHDH jobs list: https://prow.ci.openshift.org/configured-jobs/redhat-developer/rhdh
Overlay jobs list: https://prow.ci.openshift.org/configured-jobs/redhat-developer/rhdh-plugin-export-overlays
Image tags: https://quay.io/repository/rhdh/rhdh-hub-rhel9?tab=tags

Related Skills

overlay: Manage the rhdh-plugin-export-overlays repository