Configuring Requests and Limits for the AtScale Containers

In Kubernetes, requests and limits refer to the amounts of CPU and memory resources a container requires (requests), and the amount of each resource it is allowed to use (limits).

If needed, you can define requests and limits for the AtScale containers in your values override file. You can do this manually, by explicitly defining values with the resources property, or by using one of AtScale's named profiles with resourcesPreset.

Configure requests and limits manually

To configure requests and limits manually, include the resources property for the container. For example, the following configures resources for the engine's gateway container:

atscale-engine:
  gateway:
    replicaCount: 3
    
    resources:
      limits:
        cpu: 3
        memory: 2Gi
      requests:
        cpu: 1500m
        memory: 1Gi

Recommended settings

The following sections provide the minimum request and limit values for running AtScale in development and production environments.

Note

These settings are the minimum requirements for running AtScale in development and production environments. Depending on your usage of AtScale, these settings may need to be changed. You can use a tool like Goldilocks to assess how your cluster is performing under the load of your system.

For additional information on Kubernetes requests and limits, refer to the Kubernetes documentation.

Development

Component	Container	CPU Request	CPU Limit	Memory Request	Memory Limit	Replicas
Entitlement	main	250m	1	500Mi	1Gi	1
Engine	certImage	100m	250m	250Mi	500Mi	1
Engine	gateway	1	2	1Gi	2Gi	1
Engine	main	4	8	16Gi	32Gi	1
SML	main	1	2	1500Mi	3Gi	1
AtScale Proxy	main	1	2	1Gi	2000Mi	1
PostgreSQL	pgpool	1	2	1Gi	2500Mi	1
PostgreSQL	postgresql	1	2	4Gi	8Gi	1
PostgreSQL	metrics	200m	200m	500Mi	500Mi	1
Redis	master	2	4	8Gi	16Gi	1
Redis	replica	2	4	8Gi	16Gi	1
Keycloak	main	2	4	8Gi	12Gi	1
OTEL	main	1	2	1Gi	1Gi	1

Production

Important

AtScale's built-in PostgreSQL instance should not be used in production. For production environments, you should use a managed solution from your cloud provider, or set up an operator-based PostgreSQL deployment. For more information, see Configuring an External PostgreSQL Instance.

Component	Container	CPU Request	CPU Limit	Memory Request	Memory Limit	Replicas
Entitlement	main	250m	1	500Mi	1Gi	3
Engine	certImage	100m	250m	250Mi	500Mi	3
Engine	gateway	1500m	3	1Gi	2Gi	3
Engine	main	8	16	32Gi	32Gi	3
SML	main	1	2	1500Mi	3Gi	3
AtScale Proxy	main	1	2	1Gi	2000Mi	3
PostgreSQL	pgpool	1	2	1Gi	2500Mi	3
PostgreSQL	postgresql	2	4	8Gi	16Gi	3
PostgreSQL	metrics	200m	200m	500Mi	500Mi	3
Redis	master	2	4	8Gi	16Gi	1
Redis	replica	2	4	8Gi	16Gi	3
Keycloak	main	2	4	8Gi	16Gi	3
OTEL	main	1	2	1Gi	2Gi	3

Configure requests and limits using named profiles

You can use resourcesPreset to size AtScale pods using named profiles instead of manually setting resources.requests and resources.limits in each service section of your values override file. This can be configured both globally and at the chart level.

Supported values

resourcesPreset supports the following values:

• none: No resources set (Kubernetes default behavior).
• poc: Minimal resources for a proof-of-concept.
• dev: Moderate resources for development.
• prod: Production-grade resources.

Precedence order

For every container, resources are resolved in the following order (first match wins):

1. Explicit resources block: If resources: is set directly on a service section, it always takes effect and nothing else is checked.
2. global.resourcesPreset: If set to anything other than none, it applies to all charts across the umbrella deployment.
3. Chart-level resourcesPreset: The per-chart fallback when no global override is active.

Configure resourcesPreset at the global level

You can set global.resourcesPreset to apply one preset to all components:

global:
  resourcesPreset: "poc"  # none | poc | dev | prod

This single value propagates to every AtScale subchart.

Override a single chart

You can also set resourcesPreset at the chart level to override the global value. For example, to override the global value for the atscale-engine chart:

global:
  resourcesPreset: "poc"
atscale-engine:
  resourcesPreset: "dev"  # engine gets dev; all other components get poc

Override a single container with explicit values

Explicit resources always take precedent over any preset:

atscale-engine:
  # Main engine container
  resources:
    requests:
      cpu: "16"
      memory: "64Gi"
    limits:
      cpu: "32"
      memory: "128Gi"

  # cert initContainer
  certImage:
    resources:
      requests:
        cpu: "200m"
        memory: "512Mi"

  # engine-init initContainer
  defaultInitContainers:
    engineInit:
      resources:
        requests:
          cpu: "2"
          memory: "4Gi"

Additional information

• AtScale System Requirements
• Creating a Values Override File
• Configuring an External PostgreSQL Instance
• Configuring Resource Quotas for the AtScale Namespace