Quartz 5

Health probe

Properties1
tagsazure, networking, load-balancer, availability

2 min read

Health probe

What it is

A health probe is a repeating check a load-balancing service uses to decide whether a backend instance should receive new traffic. For Azure Load Balancer, a probe can use TCP, HTTP, or HTTPS.

Why it exists

A VM can be running while the application inside it cannot serve customers. A probe gives the load balancer a small, repeatable signal: send new connections to healthy instances and stop selecting an instance that fails the configured check.

Key ideas

  • TCP asks “is something listening?” HTTP and HTTPS can ask a more meaningful question by requesting a configured path. For Azure Load Balancer, HTTP(S) treats a 200 response as healthy.
  • The listener and network rules must allow the probe. For Azure Load Balancer, the AzureLoadBalancer service tag permits the IPv4 probe source by default; blocking it can make a healthy VM appear down.
  • A probe changes traffic distribution; it does not repair anything. It is different from monitoring, alerts, logs, and a recovery plan.
  • Existing connections need separate thought. When one Standard Load Balancer backend becomes unhealthy, new TCP flows go to healthy backends while established TCP flows can continue until they close or time out.
  • Keep the check useful but proportionate. A “process is alive” check can miss a broken application; a check that depends on every downstream service can mark all backends down during a separate outage.

Northwind example

Northwind runs three web VMs behind a Load Balancer. Each VM responds 200 at /health only when its web service is ready for customer requests. After a bad deployment makes one VM return 500, the probe marks that VM unhealthy and new shoppers go to the other two. Maya still uses Azure Monitor to alert the team and roll back the deployment; the probe only limits new traffic to the failed VM.

Exam lens

Use a health probe when the requirement is “remove an unhealthy backend from new load-balanced traffic.” Use azure-monitor and alerts when the requirement is “detect, investigate, and notify about the failure.”

load-balancer · application-gateway · network-watcher · azure-monitor

📘 Source: Microsoft Learn — Azure Load Balancer health probes

Graph View

Backlinks