Routing kas requests in the Kubernetes Agent

This document describes how kas routes requests to concrete agentk instances. GitLab must talk to GitLab Kubernetes Agent Server (kas) to:

Each agent connects to an instance of kas and keeps an open connection. When GitLab must talk to a particular agent, a kas instance connected to this agent must be found, and the request routed to it.

System design

For an architecture overview please see

flowchart LR subgraph "Kubernetes 1" agentk1p1["agentk 1, Pod1"] agentk1p2["agentk 1, Pod2"] end subgraph "Kubernetes 2" agentk2p1["agentk 2, Pod1"] end subgraph "Kubernetes 3" agentk3p1["agentk 3, Pod1"] end subgraph kas kas1["kas 1"] kas2["kas 2"] kas3["kas 3"] end GitLab["GitLab Rails"] Redis GitLab -- "gRPC to any kas" --> kas kas1 -- register connected agents --> Redis kas2 -- register connected agents --> Redis kas1 -- lookup agent --> Redis agentk1p1 -- "gRPC" --> kas1 agentk1p2 -- "gRPC" --> kas2 agentk2p1 -- "gRPC" --> kas1 agentk3p1 -- "gRPC" --> kas2

For this architecture, this diagram shows a request to agentk 3, Pod1 for the list of pods:

sequenceDiagram GitLab->>+kas1: Get list of running<br />Pods from agentk<br />with agent_id=3 Note right of kas1: kas1 checks for<br />agent connected with agent_id=3.<br />It does not.<br />Queries Redis kas1->>+Redis: Get list of connected agents<br />with agent_id=3 Redis-->-kas1: List of connected agents<br />with agent_id=3 Note right of kas1: kas1 picks a specific agentk instance<br />to address and talks to<br />the corresponding kas instance,<br />specifying which agentk instance<br />to route the request to. kas1->>+kas2: Get the list of running Pods<br />from agentk 3, Pod1 kas2->>+agentk 3 Pod1: Get list of Pods agentk 3 Pod1->>-kas2: Get list of Pods kas2-->>-kas1: List of running Pods<br />from agentk 3, Pod1 kas1-->>-GitLab: List of running Pods<br />from agentk with agent_id=3

Each kas instance tracks the agents connected to it in Redis. For each agent, it stores a serialized protobuf object with information about the agent. When an agent disconnects, kas removes all corresponding information from Redis. For both events, kas publishes a notification to a Redis pub-sub channel.

Each agent, while logically a single entity, can have multiple replicas (multiple pods) in a cluster. kas accommodates that and records per-replica (generally per-connection) information. Each open GetConfiguration() streaming request is given a unique identifier which, combined with agent ID, identifies an agentk instance.

gRPC can keep multiple TCP connections open for a single target host. agentk only runs one GetConfiguration() streaming request. kas uses that connection, and doesn’t see idle TCP connections because they are handled by the gRPC framework.

Each kas instance provides information to Redis, so other kas instances can discover and access it.

Information is stored in Redis with an expiration time, to expire information for kas instances that become unavailable. To prevent information from expiring too quickly, kas periodically updates the expiration time for valid entries. Before terminating, kas cleans up the information it adds into Redis.

When kas must atomically update multiple data structures in Redis, it uses transactions to ensure data consistency. Grouped data items must have the same expiration time.

In addition to the existing agentk -> kas gRPC endpoint, kas exposes two new, separate gRPC endpoints for GitLab and for kas -> kas requests. Each endpoint is a separate network listener, making it easier to control network access to endpoints and allowing separate configuration for each endpoint.

Databases, like PostgreSQL, aren’t used because the data is transient, with no need to reliably persist it.

GitLab : kas external endpoint

GitLab authenticates with kas using JWT and the same shared secret used by the kas -> GitLab communication. The JWT issuer should be gitlab and the audience should be gitlab-kas.

When accessed through this endpoint, kas plays the role of request router.

If a request from GitLab comes but no connected agent can handle it, kas blocks and waits for a suitable agent to connect to it or to another kas instance. It stops waiting when the client disconnects, or when some long timeout happens, such as client timeout. kas is notified of new agent connections through a pub-sub channel to avoid frequent polling. When a suitable agent connects, kas routes the request to it.

kas : kas internal endpoint

This endpoint is an implementation detail, an internal API, and should not be used by any other system. It’s protected by JWT using a secret, shared among all kas instances. No other system must have access to this secret.

When accessed through this endpoint, kas uses the request itself to determine which agentk to send the request to. It prevents request cycles by only following the instructions in the request, rather than doing discovery. It’s the responsibility of the kas receiving the request from the external endpoint to retry and re-route requests. This method ensures a single central component for each request can determine how a request is routed, rather than distributing the decision across several kas instances.

API definitions

syntax = "proto3";

import "google/protobuf/timestamp.proto";

message KasAddress {
    string ip = 1;
    uint32 port = 2;

message ConnectedAgentInfo {
    // Agent id.
    int64 id = 1;
    // Identifies a particular agentk->kas connection. Randomly generated when agent connects.
    int64 connection_id = 2;
    string version = 3;
    string commit = 4;
    // Pod namespace.
    string pod_namespace = 5;
    // Pod name.
    string pod_name = 6;
    // When the connection was established.
    google.protobuf.Timestamp connected_at = 7;
    KasAddress kas_address = 8;
    // What else do we need?

message KasInstanceInfo {
    string version = 1;
    string commit = 2;
    KasAddress address = 3;
    // What else do we need?

message ConnectedAgentsForProjectRequest {
    int64 project_id = 1;

message ConnectedAgentsForProjectResponse {
    // There may 0 or more agents with the same id, depending on the number of running Pods.
    repeated ConnectedAgentInfo agents = 1;

message ConnectedAgentsByIdRequest {
    int64 agent_id = 1;

message ConnectedAgentsByIdResponse {
    repeated ConnectedAgentInfo agents = 1;

// API for use by GitLab.
service KasApi {
    // Connected agents for a particular configuration project.
    rpc ConnectedAgentsForProject (ConnectedAgentsForProjectRequest) returns (ConnectedAgentsForProjectResponse) {
    // Connected agents for a particular agent id.
    rpc ConnectedAgentsById (ConnectedAgentsByIdRequest) returns (ConnectedAgentsByIdResponse) {
    // Depends on the need, but here is the call from the example above.
    rpc GetPods (GetPodsRequest) returns (GetPodsResponse) {

message Pod {
    string namespace = 1;
    string name = 2;

message GetPodsRequest {
    int64 agent_id = 1;
    int64 connection_id = 2;

message GetPodsResponse {
    repeated Pod pods = 1;

// Internal API for use by kas for kas -> kas calls.
service KasInternal {
    // Depends on the need, but here is the call from the example above.
    rpc GetPods (GetPodsRequest) returns (GetPodsResponse) {