Azure function timeout/fails to complete

Question

G'day folks,

I'm having some issues with an Azure function that I'm hoping someone might be able to help with.

We have a relatively long-running process (3-4 mins) that is being triggered from a Service Bus message, and we were having issues with the function execution ending without error and then attempting to re-process. The time take for this to happen is less than all the timeout/lock duration settings we have configured. Watching the logs (log stream, for both file system and app insights) we see the last line of the previous execution, then it kicks straight into the next.

To determine whether it's service bus related, I've also tried executing the process via a blob trigger (the process uses the file as a data source anyway) but I'm seeing the same thing except I don't see the subsequent retries.

In both scenarios I don't see anything in App insights apart from the Trace records. I don't get an exception, or even a 'request' entry. (function logic is all enclosed in try/catch blocks btw)

So my question is - Is it possible to trap these scenarios so we can determine the root cause? Currently I've got nothing to go on to try and diagnose. These errors don't happen when running locally.

FWIW we've seen this issue happen during the execution of a third-party libraries (MS Graph and an OpenXMLPowerTools library) - as we're generating documents for upload into Sharepoint. Not sure if this is relevant.

Thanking you in advance,

Tim

Answer 1

May be this is because of the plan that you are using , If you're using the Consumption plan, the default timeout is 5 minutes, but you can increase it to a maximum of 10 minutes. The maximum timeout on a Premium plan is 60 minutes. You can set your timeout as long as you want if you have a dedicated App Service plan.
Also try configuring the timeout of your function app i.e by changing the value of functionTimeout in host.json of your function app.

Answer 2

You should have a look at durable functions. They allows us to have long running processes, i.e. import/export tasks. I was able to wrap a long running import process, which takes about 20 mins to run successfully.