Allow configurable timeout period for Front Door
Currently Front Door forces a 30 second timeout for backend requests. This can severely restrict the usefulness of the service in production systems. It would be great to have the timeout period configurable to allow for a longer period of time. My understanding is that the Azure Load Balancer, which sits in a similar space as Front Door, defaults to a 4 minute timeout period.
Vivek Malhotra (BING) commented
+1 for making it a configurable option. We spent lot of time debugging an issue where we felt exception is occuring at backend and in the end, it turned out to be default timeout in afd while everything worked normally (although little slower expectedly) in the backend!
Rajinder Singh commented
Any ETA on when this will GA? This is a big problem for us.
I have an Angular application running on Azure Web Apps, that anguar application make the browser hit several APIs on Azure Web Apps as well. Front Door is failing miserably with this 30 second time out. My API also runs cpmplex process that takes a little over 30 seconds sometimes, but to my surprise they aplication is failing even when the response is less than 30 seconds, probably because the firewall needs some time to process it, and front door needs some time to route it. according to my API insight the longest it took to reply was 11 seconds, but it still fails. All my calls are asynchronous, it is an Agular app. all my infrastructure is configured via powershell, please add the option to configure the time out not just via the API please make it available via Powershell.
Any update on when this will be available? We were in the process of rolling out Front Door to our application but had to roll it back as a result of the timeout limitation. Need to understand if we need to change our approach or what is the tentative release date.
Dan Strzelec commented
Is there an update on this item? This timeout issue is a show stopper for me as well. We have many requests to our API to generate large reports that take in excess of 30s to process.
This is a showstopper for us too. There is a case, in our we app, when the timeout exceeds 30s. The request is issued by a popup window where the long delay doesn't look particularly "user unfriendly". With the Front Door, the user receives an error (the Front Door service unavailability page) in the popup. A workaround could be to re-implement the client/server interaction so it became asynchronous, but, currently, such a refactoring is problematic for us.
Microsoft, please let us configure the timeout! (And not the way you let us configure the request body SIZE limit for the Application Gateway based WAF: its maximum of 128kb is "utterly" insufficient, and there is no way to avoid the check for specific URLs.) For the TIMEOUT, the maximum should be no less than an hour - for guaranteed allowance.
In some scenarios, we aren't even going to use the load balancing capability of the Front Door: just the firewall one. Alas, we can't do even that!
We are a company that's been developing accounting and planning software for about 30 years.
We've migrated our software to Azure, and are trying to implement a Windows Application Firewall to be protected from OWASP top 10 vulnerabilities. Our attempt to implement the WAF based on the Application Gateway has failed because of the request body size limit. Even using the Firewall Policy (custom rules), it is impossible to define requests that would be exempt from the size check. Our attempt to implement the WAF based on the Front Door has failed because of the request timeout limit. The failure is preventing us from the implementation of our systems on MS Azure platform *for demanding customers*. We've been able to run our software as an Azure cloud service (classic) for years.
Thank you for your attention to this issue!
We had to rollback as well because of this issue.
Brad Darnell commented
I see this was started. Any update on when this will be available? We were in the process of rolling out Front Door to our application but have now had to roll it back as a result of this issue. Need to understand if we need to pivot to a new approach.
Andrew Innes commented
You asked: "Can you explain your scenarios a bit more? Azure Front Door is different than Azure Load Balancer particularly for the fact that it is Layer 7 i.e. HTTP/HTTPS. So, waiting and not getting a feedback from the backend/origin/app server for 30 seconds for a web request is unusual."
I think you are assuming a user interaction scenario, where 30s would indeed be a worryingly long time. However there are plenty of API scenarios where those kinds of considerations aren't relevant. In our case we have a Speech-to-Text API that is used as part of various automated batch-processing workflows by different customers, where people may post videos that are multiple GBs in size. A common pattern is to post the files as a URL reference to blob storage, which of course is very quick, but our preferred logic is to immediately fetch the blob and perform some validation steps before accepting the transcription job request. It can easily take more than 30s to fetch a large blob; from our perspective that is an arbitrary limit we would not impose by choice (we would probably set it to 5 or maybe 10 minutes).
As another example of why this is useful: we have an application that uses long polling to handle event dispatch. Any application using Microsoft .NET SignalR may use long polling when web sockets aren't available (as they don't appear to be with Front Door).
Because long polling requests wait server-side, they can be expected to have multiple-minute timeouts without this being and indication of performance issues or failed requests.
We have applications at our large fortune 500 company that regularly take 1-3 minutes for a response from the back end.
I agree it would be nice to have all responses return in less than 30 seconds, however, that is not reality.
Can we please have this configurable or at least set as below to 4 minutes?
James Brantly commented
Fair point that Azure Load Balancer works at Layer 4 and Front Door at Layer 7. My point was simply that LB often acts as a "point of entry" into Azure-based HTTP(S) services similar to Front Door. Perhaps a better analogy would be Azure Application Gateway which also works at Layer 7. It also happens to have the exact feature being requested here: https://docs.microsoft.com/en-us/azure/application-gateway/application-gateway-troubleshooting-502#request-time-out
I do agree that waiting 30 seconds for a web request is unusual and undesirable, but certainly it can happen. In our particular case we have an endpoint which runs a rather complex SQL query which averages about 35-40 seconds per run. While we'd like to improve that performance, and in fact are working to do that, it's very unfortunate that we're unable to utilize Front Door at this time since a portion of our application simply fails to load at all instead of just taking a long time.
In general terms, I think use-cases where an endpoint requires a lot of up-front processing that's not streamable to a user would benefit from this being configurable.
We would like to hear more feedback on this. Can you explain your scenarios a bit more? Azure Front Door is different than Azure Load Balancer particularly for the fact that it is Layer 7 i.e. HTTP/HTTPS. So, waiting and not getting a feedback from the backend/origin/app server for 30 seconds for a web request is unusual.