Autoscaling not scaling out on High CPU

Question

Autoscaling not scaling out on High CPU

Iain Brown 1

I have an Azure App Service that I've switched from rules based auto scale to automatic scaling as I want to not have to think about when to scale in/out.

But I'm finding that I'm not scaling out even when my CPU usage is 95-100% for reasonably long periods of time, and I'm also getting out of memory exceptions. This rise in CPU usage is tied to a rise in incoming requests at peak usage times. I've set the maximum number of instances to 10, but it never scales out beyond my initial 2. I'm using P1v3 instances.

The criteria for performing a scale out are a little opaque, but I would expect this to scale. Is there a way to diagnose why it isn't scaling?

SnehaAgrawal-MSFT 22,721 Reputation points Moderator

2024-10-30T10:53:25.71+00:00
@Iain Brown Thanks for reaching here!

Suggest you to follow- Auto scale Best practices-

Ensure the maximum and minimum values are different and have an adequate margin between them

Manual scaling is reset by autoscale minimum and maximum

Always use a scale-out and scale-in rule combination that performs an increase and decrease

Choose the appropriate statistic for your diagnostics metric

Considerations for scaling threshold values for special metrics

Considerations for scaling when multiple rules are configured in a profile

Always select a safe default instance count

Configure autoscale notifications

Further to highlight-

All thresholds are calculated at an instance level. An example is "scale out by one instance when average CPU > 80% when instance count is 2." It means scale-out when the average CPU across all instances is greater than 80%.

Also, Autoscale detects flapping and aborts the scale attempt. You see a log type of Flapping in this situation. If you see this log type, consider whether your thresholds are too narrow.

More details on Flapping are in https://learn.microsoft.com/en-us/azure/monitoring-and-diagnostics/insights-autoscale-best-practices#choose-the-thresholds-carefully-for-all-metric-types

Hope this helps, Let us know.
Iain Brown 1 Reputation point

2024-10-30T17:42:38.6633333+00:00

@SnehaAgrawal-MSFT thank you for your post. Unfortunately I think we're talking about two slightly different things, you seem to be talking about the rules based scaling, I'm trying to use the newer automatic scaling discussed here

https://learn.microsoft.com/en-us/azure/app-service/manage-automatic-scaling?tabs=azure-portal

1 answer

Your answer

SnehaAgrawal-MSFT 22,721 Reputation points Moderator

2024-10-30T10:53:25.71+00:00

@Iain Brown Thanks for reaching here!

Suggest you to follow- Auto scale Best practices-

Ensure the maximum and minimum values are different and have an adequate margin between them

Manual scaling is reset by autoscale minimum and maximum

Always use a scale-out and scale-in rule combination that performs an increase and decrease

Choose the appropriate statistic for your diagnostics metric

Considerations for scaling threshold values for special metrics

Considerations for scaling when multiple rules are configured in a profile

Always select a safe default instance count

Configure autoscale notifications

Further to highlight-

All thresholds are calculated at an instance level. An example is "scale out by one instance when average CPU > 80% when instance count is 2." It means scale-out when the average CPU across all instances is greater than 80%.

Also, Autoscale detects flapping and aborts the scale attempt. You see a log type of Flapping in this situation. If you see this log type, consider whether your thresholds are too narrow.

More details on Flapping are in https://learn.microsoft.com/en-us/azure/monitoring-and-diagnostics/insights-autoscale-best-practices#choose-the-thresholds-carefully-for-all-metric-types

Hope this helps, Let us know.
Iain Brown 1 Reputation point

2024-10-30T17:42:38.6633333+00:00

@SnehaAgrawal-MSFT thank you for your post. Unfortunately I think we're talking about two slightly different things, you seem to be talking about the rules based scaling, I'm trying to use the newer automatic scaling discussed here

https://learn.microsoft.com/en-us/azure/app-service/manage-automatic-scaling?tabs=azure-portal

Answer 1

@Iain Brown Sure, As its mentioned in the doc-

Automatic scaling is a feature in Azure App Service that simplifies the process of scaling your web apps and App Service Plans. Unlike the traditional Azure autoscale feature, where you need to set specific scaling rules (based on factors like CPU usage, memory, or predefined schedules), automatic scaling makes scaling decisions on its own, dynamically adjusting based on real-time needs.

So the use cases for Autoscale is Ideal for predictable workloads where traffic patterns are known (e.g., business hours vs. off-hours). You have full control over scaling behavior, which is useful if your app has consistent, known demand patterns.

where as use case for Automatic scaling- Best for applications with unpredictable traffic patterns where immediate responsiveness is critical. Automatic Scaling can handle sudden spikes more effectively by prewarming instances, which helps avoid delays from cold starts.

So the scenario you are expecting can be fulfilled using Autoscale-

Having said that reached out to you privately please response there.

Share via

Autoscaling not scaling out on High CPU

1 answer

Your answer