envoy_memory_request
Description
The envoy_memory_request parameter sets the memory request for the Envoy data-plane pods. These pods carry every request that enters the Rack through the ingress router, so the memory they reserve affects how much in-flight traffic and response data they can buffer.
This parameter applies only when router_type is set to contour. On Racks running the default nginx router it has no effect. It is available on AWS Racks.
Default Value
The default value is 256Mi. This default preserves existing behavior, so Racks that do not set it run the Envoy data plane with the same memory request as before.
Use Cases
- Raise the value for high-traffic Racks where the Envoy pods handle a large volume of concurrent connections.
- Raise the value for Racks serving large responses, so the data plane has headroom to buffer response bodies.
Setting the Parameter
$ convox rack params set envoy_memory_request=512Mi -r rackName
Setting parameters... OK
Viewing Current Configuration
$ convox rack params -r rackName
Additional Information
The value uses Kubernetes resource quantity notation, such as 256Mi or 1Gi. The request is the amount of memory reserved for each Envoy pod; the scheduler uses it to place the pods on nodes with sufficient capacity.
This parameter is clearable. Clearing it returns the Envoy data plane to the 256Mi default:
$ convox rack params set envoy_memory_request= -r rackName
Setting parameters... OK
Because this parameter only takes effect when router_type=contour, setting it on a Rack still running nginx records the value but does not change any running pods until you switch the Rack to Contour.
See Also
