Compare the scenarios¶
The source repository intentionally contrasts configuration choices that can affect local writes and telemetry volume. The scenario labels describe the purpose of the learning exercise, not a guarantee of a particular result for every function app.
flowchart LR
A[Same function workload] --> B[Writable scenario]
A --> C[Mounted-package scenario]
B --> D[Measure temp activity, telemetry, and response behavior]
C --> D
D --> E[Compare with equivalent load and observation window]
Source configuration comparison¶
| Concern | Writable learning scenario | Mounted-package learning scenario |
|---|---|---|
| Deployment setting | WEBSITE_RUN_FROM_PACKAGE=0 |
WEBSITE_RUN_FROM_PACKAGE=1 |
| App content behavior | Extracted application files are writable. | Package is mounted read-only. |
| Diagnostics | Detailed diagnostics enabled in the source setup. | Detailed diagnostics disabled in the source setup. |
| Default host log level | Information |
Warning |
| Application Insights sampling | 100 percent in the source example. |
5 percent in the source example. |
| Other source tuning | Temp access is enabled and SCM separation is retained. | Temp access is disabled; the example adds health checks and pre-warmed instances. |
Warning
Do not blindly transfer these app settings to a production workload. Logging, sampling, diagnostics, health checks, deployment methods, and SCM configuration are workload-specific trade-offs. Verify the setting's current support and behavior in Microsoft documentation before adopting it.
Run a fair comparison¶
- Use equivalent application code, dependency versions, regions, plans, and load profiles.
- Change a small, documented set of settings between test groups.
- Capture baseline disk, memory, CPU, latency, errors, and telemetry volume before load begins.
- Apply sustained and burst workloads, including failure paths that create or abandon temporary data.
- Compare behavior over an agreed observation window and record configuration, timestamps, and platform changes.
- Review results with the application's reliability, security, and cost owners before promoting a pattern.