Set scaling rules in Azure Container Apps

Article
08/12/2024

Azure Container Apps manages automatic horizontal scaling through a set of declarative scaling rules. As a container app revision scales out, new instances of the revision are created on-demand. These instances are known as replicas.

Adding or editing scaling rules creates a new revision of your container app. A revision is an immutable snapshot of your container app. To learn which types of changes trigger a new revision, see revision change types.

Event-driven Container Apps jobs use scaling rules to trigger executions based on events.

Scale definition

Scaling is the combination of limits, rules, and behavior.

Limits define the minimum and maximum possible number of replicas per revision as your container app scales.

Scale limit	Default value	Min value	Max value
Minimum number of replicas per revision	0	0	Maximum replicas configurable are 1,000.
Maximum number of replicas per revision	10	1	Maximum replicas configurable are 1,000.

Rules are the criteria used by Container Apps to decide when to add or remove replicas.

Scale rules are implemented as HTTP, TCP (Transmission Control Protocol), or custom.
Behavior is the combination of rules and limits to determine scale decisions over time.

Scale behavior explains how scale decisions are made.

As you define your scaling rules, it's important to consider the following items:

You aren't billed usage charges if your container app scales to zero.
Replicas that aren't processing, but remain in memory might be billed at a lower "idle" rate. For more information, see Billing.
If you want to ensure that an instance of your revision is always running, set the minimum number of replicas to 1 or higher.

Scale rules

Scaling is driven by three different categories of triggers:

HTTP: Based on the number of concurrent HTTP requests to your revision.
TCP: Based on the number of concurrent TCP connections to your revision.
Custom: Based on CPU, memory, or supported event-driven data sources such as:
- Azure Service Bus
- Azure Event Hubs
- Apache Kafka
- Redis

If you define more than one scale rule, the container app begins to scale once the first condition of any rules is met.

HTTP

With an HTTP scaling rule, you have control over the threshold of concurrent HTTP requests that determines how your container app revision scales. Every 15 seconds, the number of concurrent requests is calculated as the number of requests in the past 15 seconds divided by 15. Container Apps jobs don't support HTTP scaling rules.

In the following example, the revision scales out up to five replicas and can scale in to zero. The scaling property is set to 100 concurrent requests per second.

Example

The http section defines an HTTP scale rule.

Scale property	Description	Default value	Min value	Max value
`concurrentRequests`	When the number of HTTP requests exceeds this value, then another replica is added. Replicas continue to add to the pool up to the `maxReplicas` amount.	10	1	n/a

{
  ...
  "resources": {
    ...
    "properties": {
      ...
      "template": {
        ...
        "scale": {
          "minReplicas": 0,
          "maxReplicas": 5,
          "rules": [{
            "name": "http-rule",
            "http": {
              "metadata": {
                "concurrentRequests": "100"
              }
            }
          }]
        }
      }
    }
  }
}

Note

Set the properties.configuration.activeRevisionsMode property of the container app to single, when using non-HTTP event scale rules.

Define an HTTP scale rule using the --scale-rule-http-concurrency parameter in the create or update commands.

CLI parameter	Description	Default value	Min value	Max value
`--scale-rule-http-concurrency`	When the number of concurrent HTTP requests exceeds this value, then another replica is added. Replicas continue to add to the pool up to the `max-replicas` amount.	10	1	n/a

az containerapp create \
  --name <CONTAINER_APP_NAME> \
  --resource-group <RESOURCE_GROUP> \
  --environment <ENVIRONMENT_NAME> \
  --image <CONTAINER_IMAGE_LOCATION>
  --min-replicas 0 \
  --max-replicas 5 \
  --scale-rule-name azure-http-rule \
  --scale-rule-type http \
  --scale-rule-http-concurrency 100

Go to your container app in the Azure portal
Select Scale.
Select Edit and deploy.
Select the Scale tab.
Select the minimum and maximum replica range.
Select Add.
In the Rule name box, enter a rule name.
From the Type dropdown, select HTTP Scaling.
In the Concurrent requests box, enter your desired number of concurrent requests for your container app.

TCP

With a TCP scaling rule, you have control over the threshold of concurrent TCP connections that determines how your app scales. Every 15 seconds, the number of concurrent connections is calculated as the number of connections in the past 15 seconds divided by 15. Container Apps jobs don't support TCP scaling rules.

In the following example, the container app revision scales out up to five replicas and can scale in to zero. The scaling threshold is set to 100 concurrent connections per second.

Example

The tcp section defines a TCP scale rule.

Scale property	Description	Default value	Min value	Max value
`concurrentConnections`	When the number of concurrent TCP connections exceeds this value, then another replica is added. Replicas continue to be added up to the `maxReplicas` amount as the number of concurrent connections increase.	10	1	n/a

{
  ...
  "resources": {
    ...
    "properties": {
      ...
      "template": {
        ...
        "scale": {
          "minReplicas": 0,
          "maxReplicas": 5,
          "rules": [{
            "name": "tcp-rule",
            "tcp": {
              "metadata": {
                "concurrentConnections": "100"
              }
            }
          }]
        }
      }
    }
  }
}

Define a TCP scale rule using the --scale-rule-tcp-concurrency parameter in the create or update commands.

CLI parameter	Description	Default value	Min value	Max value
`--scale-rule-tcp-concurrency`	When the number of concurrent TCP connections exceeds this value, then another replica is added. Replicas continue to be added up to the `max-replicas` amount as the number of concurrent connections increase.	10	1	n/a

az containerapp create \
  --name <CONTAINER_APP_NAME> \
  --resource-group <RESOURCE_GROUP> \
  --environment <ENVIRONMENT_NAME> \
  --image <CONTAINER_IMAGE_LOCATION>
  --min-replicas 0 \
  --max-replicas 5 \
  --transport tcp \
  --ingress <external/internal> \
  --target-port <CONTAINER_TARGET_PORT> \
  --scale-rule-name azure-tcp-rule \
  --scale-rule-type tcp \
  --scale-rule-tcp-concurrency 100

Not supported in the Azure portal. Use the Azure CLI or Azure Resource Manager to configure a TCP scale rule.

Custom

You can create a custom Container Apps scaling rule based on any ScaledObject-based KEDA scaler with these defaults:

Defaults	Seconds
Polling interval	30
Cool down period	300

For event-driven Container Apps jobs, you can create a custom scaling rule based on any ScaledJob-based KEDA scalers.

The following example demonstrates how to create a custom scale rule.

Example

This example shows how to convert an Azure Service Bus scaler to a Container Apps scale rule, but you use the same process for any other ScaledObject-based KEDA scaler specification.

For authentication, KEDA scaler authentication parameters take Container Apps secrets or managed identity.

The following procedure shows you how to convert a KEDA scaler to a Container App scale rule. This snippet is an excerpt of an ARM template to show you where each section fits in context of the overall template.

{
  ...
  "resources": {
    ...
    "properties": {
      ...
      "configuration": {
        ...
        "secrets": [
          {
            "name": "<NAME>",
            "value": "<VALUE>"
          }
        ]
      },
      "template": {
        ...
        "scale": {
          "minReplicas": 0,
          "maxReplicas": 5,
          "rules": [
            {
              "name": "<RULE_NAME>",
              "custom": {
                "metadata": {
                  ...
                },
                "auth": [
                  {
                    "secretRef": "<NAME>",
                    "triggerParameter": "<PARAMETER>"
                  }
                ]
              }
            }
          ]
        }
      }
    }
  }
}

Refer to this excerpt for context on how the below examples fit in the ARM template.

First, you define the type and metadata of the scale rule.

From the KEDA scaler specification, find the type value.

triggers:
- type: azure-servicebus
  metadata:
    queueName: my-queue
    namespace: service-bus-namespace
    messageCount: "5"

In the ARM template, enter the scaler type value into the custom.type property of the scale rule.

...
"rules": [
  {
    "name": "azure-servicebus-queue-rule",
    "custom": {
      "type": "azure-servicebus",
      "metadata": {
        "queueName": "my-queue",
        "namespace": "service-bus-namespace",
        "messageCount": "5"
      }
    }
  }
]
...

From the KEDA scaler specification, find the metadata values.

triggers:
- type: azure-servicebus
  metadata:
    queueName: my-queue
    namespace: service-bus-namespace
    messageCount: "5"

In the ARM template, add all metadata values to the custom.metadata section of the scale rule.

...
"rules": [
  {
    "name": "azure-servicebus-queue-rule",
    "custom": {
      "type": "azure-servicebus",
      "metadata": {
        "queueName": "my-queue",
        "namespace": "service-bus-namespace",
        "messageCount": "5"
      }
    }
  }
]
...

Authentication

Container Apps scale rules support secrets-based authentication. Scale rules for Azure resources, including Azure Queue Storage, Azure Service Bus, and Azure Event Hubs, also support managed identity. Where possible, use managed identity authentication to avoid storing secrets within the app.

Use secrets

To use secrets for authentication, you need to create a secret in the container app's secrets array. The secret value is used in the auth array of the scale rule.

KEDA scalers can use secrets in a TriggerAuthentication that is referenced by the authenticationRef property. You can map the TriggerAuthentication object to the Container Apps scale rule.

Find the TriggerAuthentication object referenced by the KEDA ScaledObject specification.

In the TriggerAuthentication object, find each secretTargetRef and its associated secret.

apiVersion: v1
kind: Secret
metadata:
  name: my-secrets
  namespace: my-project
type: Opaque
data:
  connection-string-secret: <SERVICE_BUS_CONNECTION_STRING>
---
apiVersion: keda.sh/v1alpha1
kind: TriggerAuthentication
metadata:
  name: azure-servicebus-auth
spec:
  secretTargetRef:
  - parameter: connection
    name: my-secrets
    key: connection-string-secret
---
apiVersion: keda.sh/v1alpha1
kind: ScaledObject
metadata:
  name: azure-servicebus-queue-rule
  namespace: default
spec:
  scaleTargetRef:
    name: my-scale-target
  triggers:
  - type: azure-servicebus
    metadata:
      queueName: my-queue
      namespace: service-bus-namespace
      messageCount: "5"
    authenticationRef:
        name: azure-servicebus-auth

Add an entry to the auth array of the scale rule.
1. Set the value of the triggerParameter property to the value of the secretTargetRef's parameter property.
2. Set the value of the secretRef property to the name of the secretTargetRef's key property.

{
  ...
  "resources": {
    ...
    "properties": {
      ...
      "configuration": {
        ...
        "secrets": [
          {
            "name": "connection-string-secret",
            "value": "<SERVICE_BUS_CONNECTION_STRING>"
          }
        ]
      },
      "template": {
        ...
        "scale": {
          "minReplicas": 0,
          "maxReplicas": 5,
          "rules": [
            {
              "name": "azure-servicebus-queue-rule",
              "custom": {
                "type": "azure-servicebus",
                "metadata": {
                  "queueName": "my-queue",
                  "namespace": "service-bus-namespace",
                  "messageCount": "5"
                },
                "auth": [
                  {
                    "secretRef": "connection-string-secret",
                    "triggerParameter": "connection"
                  }
                ]
              }
            }
          ]
        }
      }
    }
  }
}

Some scalers support metadata with the FromEnv suffix to reference a value in an environment variable. Container Apps looks at the first container listed in the ARM template for the environment variable.

Refer to the considerations section for more security related information.

Using managed identity

Container Apps scale rules can use managed identity to authenticate with Azure services. The following ARM template passes in system-based managed identity to authenticate for an Azure Queue scaler.

"scale": {
  "minReplicas": 0,
  "maxReplicas": 4,
  "rules": [
    {
      "name": "azure-queue",
      "custom": {
        "type": "azure-queue",
        "metadata": {
          "accountName": "apptest123",
          "queueName": "queue1",
          "queueLength": "1"
        },
        "identity": "system"
      }
    }
  ]
}

To learn more about using managed identity with scale rules, see Managed identity.

From the KEDA scaler specification, find the type value.

triggers:
- type: azure-servicebus
  metadata:
    queueName: my-queue
    namespace: service-bus-namespace
    messageCount: "5"

In the CLI command, set the --scale-rule-type parameter to the specification type value.

az containerapp create \
  --name <CONTAINER_APP_NAME> \
  --resource-group <RESOURCE_GROUP> \
  --environment <ENVIRONMENT_NAME> \
  --image <CONTAINER_IMAGE_LOCATION>
  --min-replicas 0 \
  --max-replicas 5 \
  --secrets "connection-string-secret=<SERVICE_BUS_CONNECTION_STRING>" \
  --scale-rule-name azure-servicebus-queue-rule \
  --scale-rule-type azure-servicebus \
  --scale-rule-metadata "queueName=my-queue" \
                        "namespace=service-bus-namespace" \
                        "messageCount=5" \
  --scale-rule-auth "connection=connection-string-secret"

From the KEDA scaler specification, find the metadata values.

triggers:
- type: azure-servicebus
  metadata:
    queueName: my-queue
    namespace: service-bus-namespace
    messageCount: "5"

In the CLI command, set the --scale-rule-metadata parameter to the metadata values.

You need to transform the values from a YAML format to a key/value pair for use on the command line. Separate each key/value pair with a space.

az containerapp create \
  --name <CONTAINER_APP_NAME> \
  --resource-group <RESOURCE_GROUP> \
  --environment <ENVIRONMENT_NAME> \
  --image <CONTAINER_IMAGE_LOCATION>
  --min-replicas 0 \
  --max-replicas 5 \
  --secrets "connection-string-secret=<SERVICE_BUS_CONNECTION_STRING>" \
  --scale-rule-name azure-servicebus-queue-rule \
  --scale-rule-type azure-servicebus \
  --scale-rule-metadata "queueName=my-queue" \
                        "namespace=service-bus-namespace" \
                        "messageCount=5" \
  --scale-rule-auth "connection=connection-string-secret"

Authentication

Use secrets

To configure secrets-based authentication for a Container Apps scale rule, you configure the secrets in the container app and reference them in the scale rule.

A KEDA scaler supports secrets in a TriggerAuthentication which the authenticationRef property uses for reference. You can map the TriggerAuthentication object to the Container Apps scale rule.

Find the TriggerAuthentication object referenced by the KEDA ScaledObject specification. Identify each secretTargetRef of the TriggerAuthentication object.

apiVersion: v1
kind: Secret
metadata:
  name: my-secrets
  namespace: my-project
type: Opaque
data:
  connection-string-secret: <SERVICE_BUS_CONNECTION_STRING>
---
apiVersion: keda.sh/v1alpha1
kind: TriggerAuthentication
metadata:
  name: azure-servicebus-auth
spec:
  secretTargetRef:
  - parameter: connection
    name: my-secrets
    key: connection-string-secret
---
apiVersion: keda.sh/v1alpha1
kind: ScaledObject
metadata:
  name: azure-servicebus-queue-rule
  namespace: default
spec:
  scaleTargetRef:
    name: my-scale-target
  triggers:
  - type: azure-servicebus
    metadata:
      queueName: my-queue
      namespace: service-bus-namespace
      messageCount: "5"
    authenticationRef:
        name: azure-servicebus-auth

In your container app, create the secrets that match the secretTargetRef properties.
In the CLI command, set parameters for each secretTargetRef entry.
1. Create a secret entry with the --secrets parameter. If there are multiple secrets, separate them with a space.
2. Create an authentication entry with the --scale-rule-auth parameter. If there are multiple entries, separate them with a space.

az containerapp create \
  --name <CONTAINER_APP_NAME> \
  --resource-group <RESOURCE_GROUP> \
  --environment <ENVIRONMENT_NAME> \
  --image <CONTAINER_IMAGE_LOCATION>
  --min-replicas 0 \
  --max-replicas 5 \
  --secrets "connection-string-secret=<SERVICE_BUS_CONNECTION_STRING>" \
  --scale-rule-name azure-servicebus-queue-rule \
  --scale-rule-type azure-servicebus \
  --scale-rule-metadata "queueName=my-queue" \
                        "namespace=service-bus-namespace" \
                        "messageCount=5" \
  --scale-rule-auth "connection=connection-string-secret"

Using managed identity

Container Apps scale rules can use managed identity to authenticate with Azure services. The following command creates a container app with a user-assigned managed identity and uses it to authenticate for an Azure Queue scaler.

az containerapp create \
  --resource-group <RESOURCE_GROUP> \
  --name <APP_NAME> \
  --environment <ENVIRONMENT_ID> \
  --user-assigned <USER_ASSIGNED_IDENTITY_ID> \
  --scale-rule-name azure-queue \
  --scale-rule-type azure-queue \
  --scale-rule-metadata "accountName=<AZURE_STORAGE_ACCOUNT_NAME>" "queueName=queue1" "queueLength=1" \
  --scale-rule-identity <USER_ASSIGNED_IDENTITY_ID>

Replace placeholders with your values.

Go to your container app in the Azure portal.
Select Scale.
Select Edit and deploy.
Select the Scale and replicas tab.
Select the minimum and maximum replica range.
Select Add.
In the Rule name box, enter a rule name.
From the Type dropdown, select Custom.

From the KEDA scaler specification, find the type value.

triggers:
- type: azure-servicebus
  metadata:
    queueName: my-queue
    namespace: service-bus-namespace
    messageCount: "5"

In the Custom rule type box, enter the scaler type value.

From the KEDA scaler specification, find the metadata values.

triggers:
- type: azure-servicebus
  metadata:
    queueName: my-queue
    namespace: service-bus-namespace
    messageCount: "5"

In the portal, find the Metadata section and select Add. Enter the name and value for each item in the KEDA ScaledObject specification metadata section.

Authentication

Use secrets

In your container app, create the secrets that you want to reference.

Find the TriggerAuthentication object referenced by the KEDA ScaledObject specification. Identify each secretTargetRef of the TriggerAuthentication object.

apiVersion: v1
kind: Secret
metadata:
  name: my-secrets
  namespace: my-project
type: Opaque
data:
  connection-string-secret: <SERVICE_BUS_CONNECTION_STRING>
---
apiVersion: keda.sh/v1alpha1
kind: TriggerAuthentication
metadata:
  name: azure-servicebus-auth
spec:
  secretTargetRef:
  - parameter: connection
    name: my-secrets
    key: connection-string-secret
---
apiVersion: keda.sh/v1alpha1
kind: ScaledObject
metadata:
  name: azure-servicebus-queue-rule
  namespace: default
spec:
  scaleTargetRef:
    name: my-scale-target
  triggers:
  - type: azure-servicebus
    metadata:
      queueName: my-queue
      namespace: service-bus-namespace
      messageCount: "5"
    authenticationRef:
        name: azure-servicebus-auth

In the Authentication section, select Add to create an entry for each KEDA secretTargetRef parameter.

Using managed identity

Managed identity authentication is not supported in the Azure portal. Use the Azure CLI or Azure Resource Manager to authenticate using managed identity.

Default scale rule

If you don't create a scale rule, the default scale rule is applied to your container app.

Trigger	Min replicas	Max replicas
HTTP	0	10

Important

Make sure you create a scale rule or set minReplicas to 1 or more if you don't enable ingress. If ingress is disabled and you don't define a minReplicas or a custom scale rule, then your container app will scale to zero and have no way of starting back up.

Scale behavior

Scaling behavior has the following defaults:

Parameter	Value
Polling interval	30 seconds
Cool down period	300 seconds
Scale up stabilization window	0 seconds
Scale down stabilization window	300 seconds
Scale up step	1, 4, 100% of current
Scale down step	100% of current
Scaling algorithm	`desiredReplicas = ceil(currentMetricValue / targetMetricValue)`

Polling interval is how frequently event sources are queried by KEDA. This value doesn't apply to HTTP and TCP scale rules.
Cool down period is how long after the last event was observed before the application scales down to its minimum replica count.
Scale up stabilization window is how long to wait before performing a scale up decision once scale up conditions were met.
Scale down stabilization window is how long to wait before performing a scale down decision once scale down conditions were met.
Scale up step is the rate new instances are added at. It starts with 1, 4, 8, 16, 32, ... up to the configured maximum replica count.
Scale down step is the rate at which replicas are removed. By default 100% of replicas that need to shut down are removed.
Scaling algorithm is the formula used to calculate the current desired number of replicas.

Example

For the following scale rule:

"minReplicas": 0,
"maxReplicas": 20,
"rules": [
  {
    "name": "azure-servicebus-queue-rule",
    "custom": {
      "type": "azure-servicebus",
      "metadata": {
        "queueName": "my-queue",
        "namespace": "service-bus-namespace",
        "messageCount": "5"
      }
    }
  }
]

As your app scales out, KEDA starts with an empty queue and performs the following steps:

Check my-queue every 30 seconds.
If the queue length equals 0, go back to (1).
If the queue length is > 0, scale the app to 1.
If the queue length is 50, calculate desiredReplicas = ceil(50/5) = 10.
Scale app to min(maxReplicaCount, desiredReplicas, max(4, 2*currentReplicaCount))
Go back to (1).

If the app was scaled to the maximum replica count of 20, scaling goes through the same previous steps. Scale down only happens if the condition was satisfied for 300 seconds (scale down stabilization window). Once the queue length is 0, KEDA waits for 300 seconds (cool down period) before scaling the app to 0.

Considerations

In "multiple revisions" mode, adding a new scale trigger creates a new revision of your application but your old revision remains available with the old scale rules. Use the Revision management page to manage traffic allocations.
No usage charges are incurred when an application scales to zero. For more pricing information, see Billing in Azure Container Apps.
You need to enable data protection for all .NET apps on Azure Container Apps. See Deploying and scaling an ASP.NET Core app on Azure Container Apps for details.

Known limitations

Vertical scaling isn't supported.
Replica quantities are a target amount, not a guarantee.
If you're using Dapr actors to manage states, you should keep in mind that scaling to zero isn't supported. Dapr uses virtual actors to manage asynchronous calls, which means their in-memory representation isn't tied to their identity or lifetime.

Next steps

Manage secrets

Set scaling rules in Azure Container Apps

Scale definition

Scale rules

HTTP

Example

TCP

Example

Custom

Example

Authentication

Use secrets

Using managed identity

Authentication

Use secrets

Using managed identity

Authentication

Use secrets

Using managed identity

Default scale rule

Scale behavior

Example

Considerations

Known limitations

Next steps

Additional resources