HAProxy Health Checks Guide
HTTP, TCP, and agent health check types.
HAProxy Health Checks: Beyond Basic Pings
Health checks determine whether HAProxy sends traffic to a backend server. Getting them right is the difference between seamless failover and serving errors to users during incidents.
Health Check Types
| Type | How It Works | Best For |
|---|---|---|
| TCP | Connects to the port | Basic "is the process running?" check |
| HTTP | Sends an HTTP request, checks response code | Application-level health verification |
| Agent | External agent reports health status | Complex health logic, gradual drain |
HTTP Health Checks
backend web_servers
option httpchk GET /health HTTP/1.1\r\nHost:\ example.com
http-check expect status 200
# Check every 5 seconds, mark down after 3 failures, up after 2 successes
server web1 10.0.1.10:8080 check inter 5s fall 3 rise 2
server web2 10.0.1.11:8080 check inter 5s fall 3 rise 2Advanced: Content-Based Health Checks
# Check that the response body contains "healthy"
http-check expect rstring healthy
# Check that response includes database connectivity
http-check expect string "db_connected":trueAgent Health Checks
An agent check connects to a separate port on the backend server. The agent responds with a weight (0-100%) or status (up/down/drain). This enables graceful draining during deployments:
server web1 10.0.1.10:8080 check agent-check agent-port 9999 agent-inter 3sDesigning Good Health Endpoints
Your /health endpoint should check:
- Application process is running (always true if endpoint responds)
- Database connectivity (can execute a simple query)
- Critical dependency availability (cache, message queue)
- Disk space is sufficient
Important: Don't make health checks too strict. If a non-critical dependency is down, the server may still be able to serve most requests. Use separate /ready and /live endpoints for different check levels.
Eazy SaaS Tip: Use HTTP health checks that verify your application's critical path. A TCP check only confirms the port is open — the application could be in a crash loop returning 500 errors and HAProxy wouldn't know.