none
Azure Front Door - Healthcheck frequency RRS feed

  • Question

  • Firstly: I'm not sure if this is the correct place to submit this question - but I couldn't seem to find an Azure Front Door forum, and seeing as the Az CLI has the Front Door commands under the "network" namespace, I thought this was fitting.

    I have been playing with Azure Front Door to front my multi-region service (using Azure Traffic Manager for geo routing). That part of it seems to be working totally fine, however I started to notice that the Live Metrics Stream in Application Insights is reporting that my backend has around 6 "servers" with no requests being logged. The backend is an Azure Function App (using the Consumption plan).

    I thought to change the default healthcheck endpoint (/) in Azure Front Door (as I know from prior experience with Azure Function Apps that the default homepage doesn't get logged anywhere anyway). It was just a hunch at this point - that the healthcheck was causing the service to scale out so much. Once I changed the healthcheck endpoint to a generic "ping" endpoint in my Function App, I started noticing the request count started being populated in Application Insights.

    Currently, Application Insights is reporting that there are 30-40 requests per minute to my app. All to the "ping" endpoint. The Azure Front Door healthcheck interval is set to 150 (seconds) (aka 2.5 minutes), so I'm not sure why it would be hitting the endpoint so frequently.

    Can somebody tell me whether this is expected behaviour? I understand that Azure Front Door does some smart stuff around latency to the backend etc.. but this seems like a bug.

    Thanks,
    Jacob Marshall

    Tuesday, February 5, 2019 10:57 PM

Answers

  • Hi Jmshal, 

    Thank you for reaching out. 

    Azure FrontDoor, has "N" POP servers present throughput the world. The routing in Azure FrontDoor works by the intelligent health Probe. When you add a sever to a backend pool of the Azure FrontDoor, all POP servers starts probing your resources in backed pool, so when a actual request comes, based on its metrics it chooses a backend pool resource which is efficient.

    In your case, as you are using a serverless resource in the backend pool, if you think there are more request hitting, you can increase the healthcheck interval. 

    -----------------------------------------------------------------------------------------------------------

    If you found this post helpful, please give it a "Helpful" vote. 

    Please remember to mark the replies as answers if they help.

    Wednesday, February 6, 2019 4:01 AM
    Moderator

All replies

  • Hi Jmshal, 

    Thank you for reaching out. 

    Azure FrontDoor, has "N" POP servers present throughput the world. The routing in Azure FrontDoor works by the intelligent health Probe. When you add a sever to a backend pool of the Azure FrontDoor, all POP servers starts probing your resources in backed pool, so when a actual request comes, based on its metrics it chooses a backend pool resource which is efficient.

    In your case, as you are using a serverless resource in the backend pool, if you think there are more request hitting, you can increase the healthcheck interval. 

    -----------------------------------------------------------------------------------------------------------

    If you found this post helpful, please give it a "Helpful" vote. 

    Please remember to mark the replies as answers if they help.

    Wednesday, February 6, 2019 4:01 AM
    Moderator
  • Thanks for your response, msrini.

    It probably makes sense to disable this "intelligent" health probe if there is only one backend pool. Or at least have the option to disable it. Azure Traffic Manager is already performing these checks for me, but at a much more reasonable pace.

    Is there a place I can submit issues/feature requests for Azure Front Door?

    Wednesday, February 6, 2019 4:29 AM
  • You can post your feedback here: https://feedback.azure.com/forums/217313-networking?category_id=345025

    If you think this answers your question, please mark this issue as answered. 

    Wednesday, February 6, 2019 4:52 AM
    Moderator