Docker Swarm集群部署脚本需求：预生产集群复刻至生产环境

Hey there! I’ve helped plenty of teams automate Docker Swarm deployments for production, so let’s get your 30-node cluster (7 managers, 23 workers) set up without any manual docker swarm init or node join commands. Here’s a robust script-based approach tailored to your needs:

Prerequisites First #

Before running the script, make sure these boxes are checked:

• All 30 servers have matching Docker versions installed (critical for compatibility).
• Network ports required for Swarm are open: 2377/tcp (cluster management), 7946/tcp/udp (node communication), 4789/udp (overlay networking).
• You have a "control node" (can be one of your manager servers or a separate machine) that can SSH into all 30 servers without a password (set up SSH key-based auth first).

The Automation Script #

Save this as deploy-swarm-cluster.sh, update the node lists with your actual server IPs/hostnames, then run it from your control node.

#!/bin/bash

# --------------------------
# Configuration - EDIT THESE
# --------------------------
# List of your 7 manager nodes (IP or hostname)
MANAGER_NODES=(
  "prod-manager-01"
  "prod-manager-02"
  "prod-manager-03"
  "prod-manager-04"
  "prod-manager-05"
  "prod-manager-06"
  "prod-manager-07"
)

# List of your 23 worker nodes (IP or hostname)
WORKER_NODES=(
  "prod-worker-01"
  "prod-worker-02"
  # ... add the remaining 21 worker nodes here
)

# --------------------------
# Swarm Initialization Logic
# --------------------------
# Pick the first manager to initialize the swarm
FIRST_MANAGER="${MANAGER_NODES[0]}"
echo "🔧 Initializing Swarm cluster on $FIRST_MANAGER..."

# Initialize swarm with the node's public/cluster-facing IP
# (Adjust the awk command if your server uses a different primary interface)
CLUSTER_IP=$(ssh "$FIRST_MANAGER" "hostname -I | awk '{print \$1}'")
ssh "$FIRST_MANAGER" "docker swarm init --advertise-addr $CLUSTER_IP"

# Fetch join tokens for managers and workers (quiet mode to get just the token)
MANAGER_TOKEN=$(ssh "$FIRST_MANAGER" "docker swarm join-token manager -q")
WORKER_TOKEN=$(ssh "$FIRST_MANAGER" "docker swarm join-token worker -q")
SWARM_ENDPOINT="$CLUSTER_IP:2377"

# --------------------------
# Add Remaining Managers
# --------------------------
echo "🤝 Adding remaining manager nodes..."
for MANAGER in "${MANAGER_NODES[@]:1}"; do
  echo "Joining $MANAGER as a manager..."
  ssh "$MANAGER" "docker swarm join --token $MANAGER_TOKEN $SWARM_ENDPOINT"
done

# --------------------------
# Add All Worker Nodes
# --------------------------
echo "👷 Adding worker nodes..."
for WORKER in "${WORKER_NODES[@]}"; do
  echo "Joining $WORKER as a worker..."
  ssh "$WORKER" "docker swarm join --token $WORKER_TOKEN $SWARM_ENDPOINT"
done

# --------------------------
# Verify Cluster Health
# --------------------------
echo "✅ Cluster setup complete! Checking node status:"
ssh "$FIRST_MANAGER" "docker node ls"

Key Notes & Enhancements #

• Error Handling: For production, add checks (e.g., if [ $? -ne 0 ]) to exit the script if any step fails (like SSH connection issues or Docker commands failing).
• Interface Specificity: If your servers have multiple network interfaces, replace hostname -I | awk '{print $1}' with the exact IP you want Swarm to use (e.g., 10.0.0.10).
• Security: Ensure your SSH keys are secured (don’t use a passphrase-less key on shared systems, or use an SSH agent).
• Idempotency: Add a check to see if a node is already part of a swarm before running docker swarm join (use docker info | grep -q 'Swarm: active').

Verify the Cluster #

After running the script, you can double-check the cluster state by running docker node ls on any manager node—you should see all 7 managers and 23 workers listed with a Ready status.

内容的提问来源于stack exchange，提问作者Hruday