Probe interval tuning

Salves 501

Hi friends,

I have a simple website that loads in max 10 seconds.

Whenever I do maintenance on this site in AppGw 1, which has the same times as the Probe intervals and the same site, it returns quickly when we upload IIS.

However, in Brazil when we perform the same procedure, AppGw 2 takes much longer to come back, something around 5 minutes.

The only difference between AppGw is the location and number of users who access it.

You can see that the time is very short and our fear is to reduce and start a series of problems due to performance between AppGw and Backend, causing integrity failures.

What is the shortest time with a safe margin that we can configure for the site to return quickly?

Thanks.

KapilAnanth-MSFT 37,406 Reputation points Microsoft Employee

2022-07-27T05:31:35.07+00:00
Hi @Salves ,

Welcome to the Microsoft Q&A Platform. Thank you for reaching out & I hope you are doing well.
I understand that you are trying to understand the probe parameters for App Gateway.

The AppGateway marking the backend healthy solely depends on the backend giving out the expected HTTP response code.
If you are making the maintenance in the backend, and one site is taking a bit longer than the other, it is quite possible that the maintenance activity may be taking more time.

Health Probe: https://learn.microsoft.com/en-us/azure/application-gateway/application-gateway-probe-overview
Health Probe Parameters: https://learn.microsoft.com/en-us/azure/application-gateway/application-gateway-create-probe-portal

Looking at your configuration,

You are already probing the backend with a low interval (more frequently)

The exact/shortest time depends upon your application and the load that each backend instance is going to handle.

We generally recommend setting the interval not lower than 30 seconds

Thanks,
Kapil.
Salves 501 Reputation points

2022-07-27T19:24:33.567+00:00

Hi anonymous user-MSFT ,

I understand what you said.

We always use the recommended one, but the same situation occurs.

We reduced it believing that it wouldn't wait that long for the backend, but it fixed it at first, then it took a while again and sometimes it's fast.

The site comes back, but AppGw shows the message that the backend did not respond and that it is waiting time.

There are some maintenances that we do that it comes back immediately and in others it shows the message that the backend did not respond and that it is waiting for time.

I even believed that it could be the performance of AppGw, because this is the main URL where everyone connects, that is, when it comes back, all connections will try to connect again. But theoretically they were already there before, so I don't know if that can interfere.

I keep thinking if I reduce it to the shortest possible time, what kind of problem could we have? Stress on AppGw checking too often and this is bad? Or any instability in the backend would it fail connections?

It was not clear to us what it represents in practice to increase or decrease the times and the amount of times he tests.

Thanks.
KapilAnanth-MSFT 37,406 Reputation points Microsoft Employee

2022-07-28T15:13:26.193+00:00
Hi @Salves ,

By the looks of it, it may require a deeper investigation on a remote session.

If you have a support plan, I request you file a support ticket.
If not, please do let us know, we will try and help you get a one-time free technical support.

To address your query,
Lowering the probe interval will lead to the below,

Since AppGw has to send more probes in the specified interval, the load on AppGw increases.

Since your application is going to be probed more frequently, there will be more load on the website.

Thanks,
Kapil

1 answer

msrini-MSFT 9,261 Reputation points Microsoft Employee

2022-07-30T04:22:45.973+00:00

Hi,

How do you access sites which are hosted in Application gateway ? Is it via the Traffic Manager mapped URL ?

How are you determining that the second site in your AppGW is taking 5 mins ? Did you tried accessing the site using AppGW's public IP directly or are you using traffic manager's URL ?

If you are using traffic manager mapped URL, then you will need to reduce the TTL setting on your Traffic Manger profile, so the client's DNS queries are not cached.

Regards,
Karthik Srinivas
Please sign in to rate this answer.

0 comments No comments
Sign in to comment

Share via

Probe interval tuning

1 answer