Azure App Service
Azure App Service is a service used to create and deploy scalable, mission-critical web apps.
8,984 questions
Why is Azure App Service autoscaling triggered even when container memory usage is low?
![](data:image/svg+xml, %3Csvg xmlns='http://www.w3.org/2000/svg' height='64' class='font-weight-bold' style='font: 600 30.11764705882353px "SegoeUI", Arial' width='64'%3E%3Ccircle fill='hsl(300.8, 14.000000000000002%, 39%)' cx='32' cy='32' r='32' /%3E%3Ctext x='50%25' y='55%25' dominant-baseline='middle' text-anchor='middle' fill='%23FFF' %3EDP%3C/text%3E%3C/svg%3E)
Doosan Pagooah
0
Reputation points
We observed that autoscaling was triggered on our App Service Plan when the memory usage for one of the instances (e.g., p1lsdlwk00128G) reached 87%, as reported in Azure Monitor. However, when we SSH'd into the container and checked memory usage using top, the process (dotnet) was using only ~32% memory.
This discrepancy raised confusion, as the container-level memory usage did not match the autoscale trigger thresholds. We understand that Azure autoscaling evaluates memory usage at the instance/App Service Plan level, not just the container.
Could you please confirm:
- Whether autoscale rules consider total instance memory or container-specific usage?
- Why
topinside the container reports significantly lower memory than Azure Metrics? - Is there a recommended way to correlate container memory metrics with autoscale behavior?
We want to ensure autoscaling actions are well aligned with actual application memory needs and avoid unnecessary instance scaling.
Thank you for your guidance.
Azure App Service
{count} votes
-
Doosan Pagooah 0 Reputation points
2025-07-04T15:06:40.35+00:00 Inside the container, top showed our app using ~32% memory. However, free -h indicates that only ~798MiB of memory is available, and swap is already in use.
-
![](data:image/svg+xml, %3Csvg xmlns='http://www.w3.org/2000/svg' height='64' class='font-weight-bold' style='font: 600 30.11764705882353px "SegoeUI", Arial' width='64'%3E%3Ccircle fill='hsl(224.00000000000003, 15%, 31%)' cx='32' cy='32' r='32' /%3E%3Ctext x='50%25' y='55%25' dominant-baseline='middle' text-anchor='middle' fill='%23FFF' %3EBN%3C/text%3E%3C/svg%3E)
Bhargavi Naragani 6,535 Reputation points Microsoft External Staff Moderator2025-07-10T06:33:12.8133333+00:00 Hi Doosan Pagooah,
It seems like you're noticing that inside the container,
topshows 32% memory used by thedotnetprocess, but Azure Monitor and Metrics show instance-level memory usage reaching 87%, triggering autoscale.free -halso indicates low available memory and swap usage.This difference stems from how Azure App Service measures memory usage at the App Service Plan instance level, while the container’s
topcommand shows memory usage inside the container only, not the host system it’s running on. So even though your app process (dotnet) consumes 32% of container memory, other memory usage on the App Service instance like other containers, background processes, OS overhead, or file cache, may be contributing to the higher total physical memory usage, triggering the autoscaling rule.Whether autoscale rules consider total instance memory or container-specific usage?
Yes, autoscaling is triggered based on total memory usage at the App Service Plan instance level, not just individual container memory usage. https://learn.microsoft.com/en-us/azure/azure-monitor/autoscale/autoscale-overview#supported-metrics
Why
topinside the container reports significantly lower memory than Azure Metrics?The
topcommand reflects only what’s happening inside the container's namespace. It doesn’t account for:- Memory consumed by other containers sharing the same instance.
- Kernel memory, buffers/cache, and swap.
- Platform overhead and shared resources used by the App Service infrastructure.
This explains the low memory usage observed inside the container, but higher memory pressure seen in Azure Monitor. https://learn.microsoft.com/en-us/azure/app-service/web-sites-monitor
Is there a recommended way to correlate container memory metrics with autoscale behavior?
To better align container behavior with autoscale decisions, consider the following:
- Use App Service diagnostics and Container Metrics to view memory usage across all containers in the instance.
- Use Log Analytics / Azure Monitor to collect and analyze detailed memory metrics per container and per instance.
- Consider enabling per-container memory limits in your Dockerfile or configuration to help isolate usage.
- Adjust autoscale rules to also include CPU or custom metrics to avoid scaling purely on memory. https://learn.microsoft.com/en-us/azure/azure-monitor/containers/container-insights-overview
https://learn.microsoft.com/en-us/azure/app-service/monitor-instances-health-check?tabs=dotnet
Hope this helps, if you have any further concerns or queries, please feel free to reach out to us.
Sign in to comment
Sign in to answer