Backend pools and health probes in Azure - Step-by-Step Execution

Process Flow - Backend pools and health probes

Start: Client sends request

↓

Load Balancer receives request

↓

Check backend pool members

↓

Health probe checks each backend

↓

Is backend healthy?

No→Skip backend

Yes↓

Forward request to healthy backend

↓

Response sent back to client

The load balancer receives client requests, checks backend servers' health using probes, and forwards requests only to healthy servers.

Execution Sample

Azure

1. Define backend pool with 3 servers
2. Configure health probe on port 80
3. Load balancer checks health probe
4. Forward requests to healthy servers only

This setup ensures traffic goes only to servers that respond positively to health probes.

Process Table

Step	Action	Backend Server Status	Health Probe Result	Request Forwarded To
1	Load balancer receives request	Server1: Unknown, Server2: Unknown, Server3: Unknown	No probe yet	None
2	Health probe checks Server1	Server1: Healthy	Probe success	None
3	Health probe checks Server2	Server2: Unhealthy	Probe failure	None
4	Health probe checks Server3	Server3: Healthy	Probe success	None
5	Forward request to healthy backend	Server1: Healthy, Server3: Healthy	Probe success	Server1 or Server3
6	Response sent back to client	Server1: Healthy, Server3: Healthy	Probe success	Request completed

💡 Request forwarded only to servers passing health probes; unhealthy servers skipped.

Status Tracker

Variable	Start	After Step 2	After Step 3	After Step 4	Final
Server1 Status	Unknown	Healthy	Healthy	Healthy	Healthy
Server2 Status	Unknown	Unknown	Unhealthy	Unhealthy	Unhealthy
Server3 Status	Unknown	Unknown	Unknown	Healthy	Healthy
Request Forwarded To	None	None	None	None	Server1 or Server3

Key Moments - 3 Insights

Why does the load balancer skip Server2 even though it is in the backend pool?

Because the health probe failed for Server2 (see Step 3 in execution_table), marking it unhealthy, so the load balancer does not forward requests to it.

What happens if all backend servers fail the health probe?

The load balancer will not forward requests to any backend, resulting in failed or delayed responses, as no healthy servers are available (not shown in this trace but implied by the health probe logic).

Can the load balancer forward requests to multiple servers at once?

No, it forwards each request to one healthy backend server at a time, balancing load among healthy servers (Step 5).

Visual Quiz - 3 Questions

Look at the execution_table, what is Server2's status after Step 3?

A Healthy

B Unhealthy

C Unknown

D Skipped

💡 Hint

Check the 'Backend Server Status' column at Step 3 in execution_table.

At which step does the load balancer forward the request to a backend server?

A Step 2

B Step 3

C Step 5

D Step 6

💡 Hint

Look for 'Request Forwarded To' column in execution_table.

If Server3 became unhealthy, how would the 'Request Forwarded To' value change at the final step?

A Request forwarded to Server1 only

B Request forwarded to Server3 only

C Request forwarded to Server2 only

D Request forwarded to all servers

💡 Hint

Refer to variable_tracker for Server3 status and execution_table logic for forwarding.

Concept Snapshot

Backend pools group servers to receive traffic.
Health probes check if servers are healthy.
Load balancer sends requests only to healthy servers.
Unhealthy servers are skipped to avoid errors.
Health probes run regularly to update server status.

Full Transcript

When a client sends a request, the load balancer receives it and checks which backend servers are healthy using health probes. Each backend server is checked by sending a probe request, usually on a specific port like 80. If the server responds correctly, it is marked healthy; otherwise, unhealthy. The load balancer then forwards the client request only to healthy servers, skipping any unhealthy ones. This ensures reliable and efficient traffic distribution. If all servers are unhealthy, requests cannot be forwarded, causing failures. This process repeats continuously to keep backend status updated.