Why would a functions host stop for no apparent reason?

Question

I have an Azure app service that is running an eternal Durable Functions orchestration, as well as a couple of other orchestrations and several activities, on the cosumption plan. All runs rather smoothly most of the time, but on several occasions the function host stops for no apparent reason - no deployment, no changes in configuration or anything of the sort. Sometimes when that happens the app will somehow restart itself after 2-3 hours, but there have been cases where it won't come alive until manually restarted (from the Azure portal).

As a workaround to the above problem, I have tried creating a timer-triggered function in another app service that pings an HTTP function in the problematic app service, which in turn creates a new instance of the eternal orchestration. The problem with that is when the host is down, attempting to call the HTTP function gives back a 503 Service unavailable response.

I have the following coming up in diagnostics, so I am wondering what the causes for that behavior might be.

Could it be that the host gets shut down because it consumes too much CPU and is hitting a quota of sorts, although I have never been able to find documentation anywhere that such quotas might even exist?

And what would be the recommended resolution for the situation? Splitting up functions across different app services? Going for a dedicated plan?

Any advice much appreciated,

Teodora Gancheva

Answer 1

Hello,

Welcome to Microsoft Q&A,

Running Durable Functions on an Azure Consumption Plan can sometimes lead to unexpected host shutdowns, especially with long-running or resource-intensive orchestrations.

Resource Constraints on the Consumption Plan

• Memory and CPU Limits: Each instance in the Consumption Plan is typically limited to 1.5 GB of memory and one CPU. (https://learn.microsoft.com/en-us/azure/azure-functions/functions-scale)
• Impact on Long-Running Functions: Resource-intensive operations may exceed these limits, causing the host to shut down or restart unexpectedly.

Potential Fixes:

1. Upgrade to Premium Plan: The Premium Plan offers enhanced performance, VNET integration, and no idle timeouts, making it suitable for long-running functions.
2. Optimize Function Code: Review your orchestrations and activities to ensure efficient resource usage, minimizing memory and CPU consumption.

Please Upvote and Accept the Answer if it helps!!