CD Tidbit — Multi-Service Deploy (Fan-out / Fan-in)

Deploy two independent services in one pipeline run using parallel stages that converge at the end.

              ┌─── Deploy Service A (cart-frontend) ───┐
  Pipeline ───┤                                         ├──► Converge
              └─── Deploy Service B (cart-backend)  ───┘
                   (both run at the same time)           (runs only after both succeed)

---

How it works

1. Fan-out — Harness starts both Deploy stages at the same time (parallel block).
2. Each stage deploys its own service into the shared Kubernetes namespace independently.
3. Fan-in — The Converge stage starts only after both deploys succeed. If either fails, it rolls back that service and the pipeline stops.

The shared infrastructure needs allowSimultaneousDeployments: true — without it, Harness serializes parallel stages on the same infra, and they run one after another instead of together.

---

Prerequisites

What you need	Default in this repo
Kubernetes Cluster connector (with delegate)	account.k8s_delegate
DockerHub connector (anonymous OK)	account.Docker
GitHub connector pointing to this repo	account.GitHub
Harness org + project	default / project_1

---

Files

.harness/
  pipeline.yaml        ← the pipeline (fan-out → fan-in)
  serviceA.yaml        ← Service A definition (cart-frontend, nginx:1.27-alpine)
  serviceB.yaml        ← Service B definition (cart-backend, nginx:stable-alpine)
  environment.yaml     ← shared Dev environment
  infrastructure.yaml  ← shared K8s infra (allowSimultaneousDeployments: true)

k8s/
  serviceA/            ← Go-templated K8s manifests for Service A
    namespace.yaml
    deployment.yaml
    service.yaml
    values.yaml
  serviceB/            ← Go-templated K8s manifests for Service B
    namespace.yaml
    deployment.yaml
    service.yaml
    values.yaml

The services pull manifests directly from this repo via a GitHub connector. No separate upload step needed.

---

Setup

1. Replace placeholders

Search all .harness/*.yaml files for # REPLACE: and swap in your actual values:

• orgIdentifier — your Harness org
• projectIdentifier — your Harness project
• connectorRef: account.GitHub — your GitHub connector (pointing to this repo)
• connectorRef: account.k8s_delegate — your K8s connector
• connectorRef: account.Docker — your DockerHub connector
• branch: main — your default branch name
• Namespace tidbit-0068 — whatever namespace you want

2. Import entities (in this order)

1. Services — Deployments → Services → New Service → YAML — import serviceA.yaml, then serviceB.yaml
2. Environment — Deployments → Environments → New → YAML — import environment.yaml
3. Infrastructure — open env_dev_tid0068 → Infrastructure → New → YAML — import infrastructure.yaml
4. Pipeline — Pipelines → Create → Inline → YAML — paste pipeline.yaml

3. Run

Hit Run on the pipeline. Both Deploy stages should start together and finish independently. Converge runs last.

---

Verify it worked

kubectl -n tidbit-0068 get deployments
# Expected:
# cart-frontend-deployment   1/1   Ready
# cart-backend-deployment    1/1   Ready

The Converge stage logs will show:

Service A stage status: SUCCESS
Service B stage status: SUCCESS

---

Troubleshooting

Stages run one after another, not in parallel The infra is missing allowSimultaneousDeployments: true. Harness adds a resource-constraint step to every deploy stage — without this flag, two stages on the same infra are serialized.

namespace "tidbit-0068" not found namespace.yaml must be listed first in each service's paths[] array. Also confirm createNamespace: true in values.yaml.

One pod overwrites the other Each service must have a unique name in its values.yaml (cart-frontend vs cart-backend). This drives distinct Deployment names and app: label selectors.

Converge runs even though a deploy failed Check the when condition on the Converge stage — it should be stageStatus: Success (the default). If changed, restore it.

---

Learn more

• Multi-service deployments
• Run stages in parallel
• Sequential and parallel deployments