Configure and deploy Azure function apps

Bundled into the existing Developing AI Cloud Solutions on Azure premium course — no separate purchase.

14-day money-back guarantee — no questions asked.

Included in this chapter:

Hosting option	Scale-out behavior	Function timeout (default / max, minutes)	Cold-start mitigation	Deployment technology and slots
Flex Consumption	Event driven and fast, per-function scaling, to 1,000 instances; scales to zero	30 / unbounded	Always ready instances, optional, default 0	One deploy only; slots not supported, rolling updates instead
Elastic Premium	Event driven across the plan; Windows to 100, Linux 20-100 by region; never below one instance	30 / unbounded	Always ready instances plus a prewarmed buffer, default 1 prewarmed	Zip deploy; 3 slots including production
Dedicated (App Service)	Manual scale or autoscale, slower than elastic scale; no scale to zero	30 / unbounded, and unbounded requires Always on	Always on setting; the host runs continuously	Zip deploy; 1 to 20 slots
Azure Container Apps	Event driven, 300 to 1,000 replicas depending on how the app was created	30 / unbounded when minimum replicas is one or more	Minimum replicas of one or more keeps the host running	Container image; uses revisions rather than slots
Consumption (legacy)	Event driven; Windows 200, Linux 100 instances; scales to zero	5 / 10	Internal placeholder optimization; no always-ready or prewarmed setting	Zip deploy on Windows, external package URL on Linux; 2 slots

Cheat sheet

The Consumption plan scales to zero and bills per execution, at the cost of cold starts
The Premium plan pairs always ready instances with a prewarmed buffer, and adds VNet integration
Dedicated (App Service) and Flex Consumption plans trade off predictable capacity vs fast elastic scale
The Always On setting exists only on Dedicated plans, not Consumption
WEBSITE_RUN_FROM_PACKAGE runs the app from a read-only mounted package, on the plans that allow it
func azure functionapp publish deploys the app, using a remote build for Python on Linux
Deployment slots let you stage and swap a new build with warm-up and easy rollback
Application settings surface as environment variables; AzureWebJobsStorage is required by the runtime
Key Vault references let an app setting resolve a secret from Key Vault at runtime without code changes
Identity-based (managed identity) connections replace secret connection strings for triggers and bindings
Python functions run in a language worker out-of-process, pinned to a supported Functions runtime version
The Python v2 programming model defines functions with decorators in function_app.py
Python dependencies are declared in requirements.txt and restored during deployment build

Unlock with Premium — includes all practice exams and the complete study guide.

Cheat sheet