local-dev-setup

name	local-dev-setup
description	Deploy Konflux on a local Kind cluster. Supports development mode (operator outside cluster for fast iteration) and preview mode (operator deployed in-cluster from release/build). Use when the user wants to start, restart, or recreate a local Konflux environment, run the operator locally, install integrations like sigstore or quay, or troubleshoot local deployment.

Local Development Setup

Deploy Konflux on a local Kind cluster. Two modes are available:

• Development (OPERATOR_INSTALL_METHOD=none) — operator runs on the host for fast rebuild-restart cycles. This is the default mode for this skill.
• Preview (release, build, or local) — operator runs in-cluster; deploy-local.sh handles everything end-to-end.

If the user doesn't specify, use development mode. Ask only if their intent is ambiguous.

Full documentation: operator/docs/content/docs/installation/install-local.md

Prerequisites

• kind, kubectl, podman (or docker)
• helm (only for sigstore integration)
• A configured scripts/deploy-local.env file (see "Env File" section)

Env File

The env file at scripts/deploy-local.env must exist before running. If it doesn't, copy the template:

cp scripts/deploy-local.env.template scripts/deploy-local.env

Required secrets (must already be populated by the user — never generate or guess these):

Variable	Description
GITHUB_APP_ID	Numeric GitHub App ID
GITHUB_PRIVATE_KEY or GITHUB_PRIVATE_KEY_PATH	PEM key content or path to .pem file
WEBHOOK_SECRET	GitHub webhook secret

GitHub App setup guide: operator/docs/content/docs/guides/github-secrets.md

OPERATOR_INSTALL_METHOD values

Value	Mode	Description
none	Development	Operator runs on host via make run
release	Preview	Install from latest GitHub release
build	Preview	Build operator image from source
local	Preview	Kustomize from current checkout

Set the appropriate value in deploy-local.env before running. If the file already has a different value, change it using StrReplace.

(Optional) Delete and Recreate the Kind Cluster

If the user asks to recreate the cluster, or if the cluster is in a broken state, delete it first:

kind delete cluster --name konflux

The containerd image cache (~/.cache/konflux-ci/containerd-cache) persists across cluster deletions when ENABLE_IMAGE_CACHE=1, making recreation much faster. Do not clear the cache unless the user explicitly asks or there is a specific issue (e.g. corrupted images, Kubernetes version change):

rm -rf ~/.cache/konflux-ci/containerd-cache

---

Development Mode (operator outside cluster)

Use this when the user is developing the operator and needs fast iteration.

Checklist

Track progress with TodoWrite:

- [ ] Ensure scripts/deploy-local.env exists and has OPERATOR_INSTALL_METHOD=none
- [ ] (Optional) Delete existing Kind cluster
- [ ] Run deploy-local.sh to create cluster and deploy dependencies
- [ ] Install CRDs: cd operator && make install
- [ ] Run operator locally: cd operator && make run
- [ ] Apply the Konflux CR
- [ ] Wait for Konflux to become ready
- [ ] (Optional) Install integrations

Step 1: Set OPERATOR_INSTALL_METHOD=none

Verify scripts/deploy-local.env has:

OPERATOR_INSTALL_METHOD=none

Step 2: Deploy Cluster and Dependencies

./scripts/deploy-local.sh

This creates the Kind cluster and installs dependencies (Tekton, cert-manager, etc.) but skips operator installation. Takes several minutes. Use block_until_ms: 0 and poll the terminal for "Deployment Complete!" or an error.

Step 3: Install CRDs and Run the Operator

From the operator/ directory:

cd operator && make install

Then start the operator — this is a long-running process, background it immediately:

make run

Use block_until_ms: 0 for make run since it runs indefinitely. Poll the terminal to confirm startup (look for "Starting workers" or controller log lines).

Step 4: Apply the Konflux CR and Wait

In a separate shell (the operator terminal is occupied), apply the CR using the shim script. It uses scripts/resolve-konflux-cr.sh to pick the right CR (defaults to the base CR; auto-selects konflux-e2e.yaml when Quay credentials are set).

Run these commands from the repository root (use working_directory parameter pointing to the repo root, since the previous step leaves the shell in operator/):

bash skills/local-dev-setup/scripts/apply-konflux-cr.sh

To use a specific CR file:

bash skills/local-dev-setup/scripts/apply-konflux-cr.sh operator/config/samples/konflux_v1alpha1_konflux.yaml

Then wait for Konflux to become ready:

bash skills/local-dev-setup/scripts/wait-for-konflux.sh

Both scripts are in the command allowlist and run without user confirmation.

Restarting the Operator

When the user modifies operator code and wants to restart:

1. Stop the running make run process (kill it via the terminal pid)
2. Run make run again from operator/

No need to recreate the cluster or redeploy dependencies. If API types changed, run make install before make run.

---

Preview Mode (operator in cluster)

Use this when the user wants a fully self-contained Konflux instance without running the operator on the host.

Checklist

Track progress with TodoWrite:

- [ ] Ensure scripts/deploy-local.env exists with desired OPERATOR_INSTALL_METHOD (release/build/local)
- [ ] (Optional) Delete existing Kind cluster
- [ ] Run deploy-local.sh — handles everything
- [ ] Wait for Konflux to become ready
- [ ] (Optional) Install integrations

Step 1: Set OPERATOR_INSTALL_METHOD

Set the desired method in scripts/deploy-local.env:

• release — latest GitHub release (simplest, recommended for preview)
• build — builds the operator image from the current checkout
• local — applies kustomize manifests from the current checkout with the latest released image

Step 2: Run deploy-local.sh

./scripts/deploy-local.sh

The script handles everything: Kind cluster, dependencies, operator deployment, CR application, secrets, and readiness wait. Takes several minutes. Use block_until_ms: 0 and poll for "Deployment Complete!" or an error.

For Podman on macOS:

export KIND_EXPERIMENTAL_PROVIDER=podman
./scripts/deploy-local.sh

---

(Optional) Install Integrations

Integrations live in integrations/. Each has its own install script. Install them after Konflux is ready unless the integration docs say otherwise.

Sigstore

Deploys Fulcio, Rekor, CT Log, and Trillian for keyless signing. Requires helm.

./integrations/sigstore/install.sh

Patches the Konflux CR and Tekton Chains config automatically. Takes several minutes (Helm install with --wait --timeout 15m). Use block_until_ms: 0 and poll.

Quay (Self-Hosted Image Controller)

Configures image-controller to use the self-hosted Quay deployed by deploy-deps.sh. Only works if Quay was deployed (requires SKIP_QUAY=false when running deploy-deps).

./integrations/quay/configure-image-controller.sh

For external Quay (quay.io), set QUAY_TOKEN and QUAY_ORGANIZATION in deploy-local.env and use a CR that enables image-controller:

KONFLUX_CR=operator/config/samples/konflux-e2e.yaml ./scripts/deploy-local.sh

Adding New Integrations

If the user asks to install an integration not listed here, look for an install script in the relevant integrations/<name>/ directory and run it.

Verification

• Operator logs: in dev mode, check the terminal running make run; in preview mode, kubectl logs -n konflux-operator deployment/konflux-operator-controller-manager
• Konflux status: kubectl get konflux konflux -o jsonpath='{.status.conditions}'
• UI: https://localhost:9443
• Credentials: [e-mail] / password (also [e-mail])
• Registry: localhost:5001 (if ENABLE_REGISTRY_PORT=1)

Key Files

File	Purpose
scripts/deploy-local.sh	Main deploy script
scripts/deploy-local.env.template	Config template
scripts/setup-kind-local-cluster.sh	Kind cluster setup
deploy-deps.sh	Dependency installation
scripts/deploy-secrets.sh	Secret creation

Troubleshooting

Symptom	Fix
make run fails with CRD errors	Run make install first, then make run
Cluster exists but is broken	Delete and recreate: kind delete cluster --name konflux && ./scripts/deploy-local.sh
Port 5001 in use	Set REGISTRY_HOST_PORT to another port in deploy-local.env, or set ENABLE_REGISTRY_PORT=0
Operator panics on start	Check if secrets exist: kubectl get secrets -n konflux-build-service — if missing, re-run deploy-local.sh
Konflux CR not becoming Ready	kubectl get konflux konflux -o yaml and inspect .status.conditions