Azure Load balancer Health Probe

Azure Load balancer health Probe

Azure Load balancer health probe is used to detect the endpoint status in the backend pool. When a health probe fails , load balancer stops sending new connections to respective unhealthy instances. In this outbound connection is not affected , only inbound connection is affected. 

Health probe supports multiple protocols , and these protocols usage depends upon Load balancer SKU. 

Load balancer health probes originate from IP address 168.63.129.16 and these should not be blocked from probes to mark your instances as UP. 

Probe Configuration 

In Azure Load Balancer , health probe configuration consists of following components.

• Duration of Interval between individual probes. 
• Protocol
• Port
• HTTP path to use the HTTP GET when using HTTP(S) probes.  

Application signal , Signal Detection & Load Balancer Reaction. 

The interval value determines how frequently the health probe will probe for a response from your backend pool instances. If the health probe fails, it will immediately mark your backend pool instances as unhealthy. On the next healthy probe up, the health probe will immediately mark your backend pool instances as healthy.

For example, a health probe set to five seconds. The time at which a probe is sent isn't synchronized with when your application may change state. The total time it takes for your health probe to reflect your application state can fall into one of the two following scenarios:

1. If your application produces a time-out response just before the next probe arrives, the detection of the events will take 5 seconds plus the duration of the application time-out when the probe arrives. You can assume the detection to take slightly over 5 seconds.
2. If your application produces a time-out response just after the next probe arrives, the detection of the events won't begin until the probe arrives and times out, plus another 5 seconds. You can assume the detection to take just under 10 seconds.

For this example, once detection has occurred, the platform will take a small amount of time to react to the change.

The reaction depends on:

• When the application changes state
• When the change is detected
• When the next health probe is sent
• When the detection has been communicated across the platform

Assume the reaction to a time-out response will take a minimum of 5 seconds and a maximum of 10 seconds to react to the change.