Load Balancer Health Probe time

Question

Good morning, I am using a Load Balancer to distribute load between my virtual machines, to which I constantly send UDP packets, I have Health Probe rules configured to check that they have not been dropped.

My problem comes when I am sending constant packets to these machines and one of them falls, the case is that the balancer up to 10 seconds (intervals of 2 x 5 seconds) is not able to verify that it should not send packets to that machine and continues sending you packages which get lost.
I would like to know if there is a faster way to buy if a machine is down or not.

Thanks in advance.

Answer 1

Hi @Ángel Salar García ,

Thanks for reaching us.

The minimum probe interval is 5 seconds and the minimum number of unhealthy responses is 2, it is the default configuration on Loadbalancer health probe. You dont have option now to reduce the minimum probe interval or unhealthy response. So when you have VM behind the LB and if the backend is down then application starts producing a time-out only after probe timeout value defined i.e (5x2 Seconds), yes for that period you see traffic will be forwarded to the backend which is down as its still not marked as unhealthy.

Health probe is the only way for LB to determine the health, no other ways it uses to determine the health of backend, If we have less probe interval time like a minute or 2 then it will be helpful in your scenario but with the existing probe configuration 5x2 seconds is minimum.

Ref: https://learn.microsoft.com/en-us/azure/load-balancer/load-balancer-custom-probe-overview#understanding-application-signal-detection-of-the-signal-and-reaction-of-the-platform

Hope this was helpful. Please let us know in case of any additional questions or concerns.

Please "Accept the answer" if the information helped you. This will help us and others in the community as well