Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

InferencePool config proposal for API review #162

Open
wants to merge 2 commits into
base: main
Choose a base branch
from

Conversation

ahg-g
Copy link
Contributor

@ahg-g ahg-g commented Jan 7, 2025

@k8s-ci-robot k8s-ci-robot added the do-not-merge/hold Indicates that a PR should not merge because someone has issued a /hold command. label Jan 7, 2025
@k8s-ci-robot
Copy link
Contributor

[APPROVALNOTIFIER] This PR is APPROVED

This pull-request has been approved by: ahg-g

The full list of commands accepted by this bot can be found here.

The pull request process is described here

Needs approval from an approver in each of these files:

Approvers can indicate their approval by writing /approve in a comment
Approvers can cancel approval by writing /approve cancel in a comment

@k8s-ci-robot k8s-ci-robot added cncf-cla: yes Indicates the PR's author has signed the CNCF CLA. approved Indicates a PR has been approved by an approver from all required OWNERS files. size/XL Denotes a PR that changes 500-999 lines, ignoring generated files. labels Jan 7, 2025
@kfswain
Copy link
Collaborator

kfswain commented Jan 7, 2025

This is intended to be post-v0.1, correct?

}

// ExtensionRef is a reference to the extension deployment.
type ExtensionRef struct {
// A selector for the pods that run the deployment.
Copy link
Contributor Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

  1. Need to support extensions possibly running outside the cluster
  2. Need to support "localhost" to run extensions as a sidecar

// A selector for the pods that run the deployment.
//
// +kubebuilder:validation:Required
Selector map[string]string `json:"selector"`
Copy link
Contributor Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

@danehans @robscott If we change this to a Service ref, then the absence of it could mean that the extension is deployed as a sidecar.

Copy link
Member

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

If this is a generic k8s object ref that defaults to Service, is there any way that we could reuse that to describe a sidecar? @danehans or @arkodg do you have any thoughts on what the ideal API surface would look like for an extension deployed as a sidecar?

Copy link

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

assuming sidecar here, is sidecar in the gateway pod, if it looks like a backendRef, you could route back to the Gateway Service and have the sidecar listen on a different port, and other implementations like Envoy Gateway could use the custom Backend Resource to backendRef to route via UDS

Copy link
Member

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Yep, sidecar on the gateway pod, that's a good idea.

Copy link
Contributor Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

ok, so we should be good with a service reference here, right?

Copy link

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

@ahg-g something that looks like https://github.com/kubernetes-sigs/gateway-api/blob/50cf145b0dd1849ebd03c5866a92797862f23efe/apis/v1/object_reference_types.go#L96 should solve the Service case and also allow for other custom backend types

@k8s-ci-robot
Copy link
Contributor

@ahg-g: The following tests failed, say /retest to rerun all failed tests or /retest-required to rerun all mandatory failed tests:

Test name Commit Details Required Rerun command
pull-gateway-api-inference-extension-verify-main 5e51daf link true /test pull-gateway-api-inference-extension-verify-main
pull-gateway-api-inference-extension-test-unit-main 5e51daf link true /test pull-gateway-api-inference-extension-test-unit-main

Full PR test history. Your PR dashboard. Please help us cut down on flakes by linking to an open issue when you hit one in your PR.

Instructions for interacting with me using PR comments are available here. If you have questions or suggestions related to my behavior, please file an issue against the kubernetes-sigs/prow repository. I understand the commands that are listed here.

Comment on lines +130 to +134

// The port number on the pods running the extension. Defaults to 9002 if not set.
//
// +kubebuilder:default=9002
TargetPortNumber *int32 `json:"targetPortNumber"`
Copy link
Contributor

@danehans danehans Jan 10, 2025

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Include omitempty if this is an optional field.

Thoughts on how to secure this connection? For example:

...
extensionRef:
  targetPortNumber: 9002
  protocol: TLS # Maybe default to HTTP with enum for TLS, UDS (e.g. sidecar extension)
  tls: # Required when protocol is TLS
      certificateRefs:
      - kind: Secret
        name: example-com

xref for additional details.

@@ -61,8 +61,100 @@ type InferencePoolSpec struct {
// +kubebuilder:validation:Maximum=65535
// +kubebuilder:validation:Required
TargetPortNumber int32 `json:"targetPortNumber"`

// EndpointPickerConfig selects and configures the endpoint picking algorithm to apply on the requests sent
Copy link
Contributor

@danehans danehans Jan 10, 2025

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

How does the term "picker" differ from "selector"? I ask b/c users have established a mental model for Endpoint Selector. Unless "picker" has a different semantic meaning than "selector", I'd like to see the project standardize on "selector".

Since EndpointPickerConfig is not optional, this means every implementation must support the EPP extension. If that is the case, the reference EPP extension must be pluggable using a standardized interface, e.g. CNI, CNI, etc.

cc: @robscott

}

// ExtensionConnection encapsulates options that configures the connection to the extension.
type ExtensionConnection struct {
Copy link
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Based on my above feedback, consider including targetPortNumber, protocol, tls, etc. in this struct since these concepts are related to connection.

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Labels
approved Indicates a PR has been approved by an approver from all required OWNERS files. cncf-cla: yes Indicates the PR's author has signed the CNCF CLA. do-not-merge/hold Indicates that a PR should not merge because someone has issued a /hold command. size/XL Denotes a PR that changes 500-999 lines, ignoring generated files.
Projects
None yet
Development

Successfully merging this pull request may close these issues.

6 participants