Files

T

HeruEdhel 59330bdf84 feat: Implement runtime API cache store and related configurations

- Added RuntimeApiCacheStore for managing API cache entries in memory.
- Introduced StorageTarget enum to define various storage targets.
- Created StoredCacheEntry record to encapsulate cache entries.
- Defined SyncState enum to represent the synchronization state of cache entries.
- Added ControlPlaneConfigurationSource enum for different configuration sources.
- Developed ControlPlaneRuntimeConfig class for runtime configuration management.
- Implemented ControlPlaneStartupConfigResolver for resolving configuration from various sources.
- Created ControlPlaneStartupSettings to hold startup configuration values.
- Added ResolvedControlPlaneStartupConfig record to encapsulate resolved configuration.
- Implemented tests for batch client and cache client functionalities.
- Added tests for ControlPlaneStartupConfigResolver to validate configuration resolution logic.

2026-06-05 07:00:14 -07:00

2.0 KiB

Raw Blame History

Batched API Calls

Batching lets multiple plugins enqueue control-plane calls that are sent as one HTTP request to /api/v1/batch.

Use batching when:

many plugins read data at the same time
looped tasks report state periodically
call latency can be handled asynchronously

Do not use batching when:

the caller needs a blocking result immediately
the endpoint should bypass queueing for operational safety

One-Off Calls

CompletableFuture<ServerProfile> future = NetworkPlugin.client()
    .batch()
    .get(ProfileEndpoint.VALUE, ProfileEndpoint.VALUE.uri(), ServerProfile.class);

CompletableFuture<JsonObject> future = NetworkPlugin.client()
    .batch()
    .enqueueJson("GET", "/api/v1/server/profile", null);

Queued calls complete when the batch loop flushes, or when code calls:

NetworkPlugin.client().batch().flush();

Looped Calls

LoopHandle handle = NetworkPlugin.client().batch().registerLoopedCall(
    LoopedApiCall.builder(ServerProfile.class)
        .id("server-profile-loop")
        .interval(Duration.ofSeconds(5))
        .method("GET")
        .uri("/api/v1/server/profile")
        .onSuccess(profile -> {
            // update local state
        })
        .onFailure(error -> {
            // log or mark degraded
        })
        .build()
);

Cancel when no longer needed:

handle.cancel();

Batch Endpoint Contract

Request:

{
  "requests": [
    {
      "id": "generated-call-id",
      "method": "GET",
      "uri": "/api/v1/server/profile",
      "body": null
    }
  ]
}

Response:

{
  "responses": [
    {
      "id": "generated-call-id",
      "status": 200,
      "body": { "data": {} },
      "error": null
    }
  ]
}

Each item completes independently. A failed item does not fail unrelated futures.

Defaults

endpoint: /api/v1/batch
flush interval: 250ms
max batch size: 100
batching enabled: true

Configure with ControlPlaneClient.builder().

2.0 KiB Raw Blame History