02-client-server

Client-Server Example

Demonstrates a TCP server and multiple clients running across Kubernetes nodes using Reaper. The server and clients communicate over the host network using socat, which is available on all Kubernetes nodes.

Topology

┌──────────────────┐     ┌──────────────────┐     ┌──────────────────┐
│  worker (server)  │     │ worker2 (client)  │     │ worker3 (client)  │
│                   │     │                   │     │                   │
│  socat listening  │◄────│  socat connects   │     │  socat connects   │
│  on port 9090     │◄──────────────────────────────│  every 5 seconds  │
└──────────────────┘     └──────────────────┘     └──────────────────┘

• Server responds with its hostname and a timestamp on each connection.
• Clients connect every 5 seconds and log the response.
• All processes run directly on the host via Reaper (no container isolation).

Setup

From the repository root:

./examples/client-server/setup.sh

This creates a 4-node Kind cluster (reaper-client-server-demo) with:

• 1 control-plane node
• 1 worker labeled role=server
• 2 workers labeled role=client
• Reaper runtime installed on all nodes
• A server-config ConfigMap containing the server node's internal IP

Prerequisites

• Docker
• kind
• kubectl
• Ansible (pip install ansible)

Running the Demo

# Start the server (waits for connections on port 9090)
kubectl apply -f examples/client-server/server-daemonset.yaml

# Start the clients (connect to the server every 5 seconds)
kubectl apply -f examples/client-server/client-daemonset.yaml

# Watch client logs — each client reports the server's response
kubectl logs -l app=demo-client --all-containers --prefix -f

Expected output:

[pod/demo-client-abc12/client] Client starting on reaper-client-server-demo-worker2, server at 172.18.0.3:9090
[pod/demo-client-abc12/client] [reaper-...-worker2] 14:32:05 <- Hello from reaper-...-worker — 14:32:05
[pod/demo-client-xyz34/client] [reaper-...-worker3] 14:32:07 <- Hello from reaper-...-worker — 14:32:07

Check the server side:

kubectl logs -l app=demo-server -f

How It Works

1. Server runs a socat TCP listener on port 9090. On each connection it responds with its hostname and the current time.
2. Clients read the server's IP from the server-config ConfigMap (injected as the SERVER_IP environment variable) and connect using socat.
3. Since Reaper uses host networking, the server listens on the node's real IP and clients connect directly to it — no Kubernetes Service or port-forward required.

Cleanup

# Remove the workloads
kubectl delete -f examples/client-server/

# Delete the Kind cluster
./examples/client-server/setup.sh --cleanup