plugin-cluster-manager 1.1.5 → 1.1.7

package/README.md

@@ -1,15 +1,27 @@
 # plugin-cluster-manager
 
 ## Overview
-Monitor cluster nodes, async tasks, workflow executions, event queues, locks, caches, and Redis metrics.
+Monitor and operate NocoBase cluster nodes, async tasks, workflow executions, event queues, locks, caches, Redis metrics, container workers, and worker package installation.
 
 ## Features
-- **Cluster Dashboard**: Realtime view of all active NocoBase nodes and workers.
-- **Queue Management**: Inspect active, pending, and failed background jobs (BullMQ).
-- **Redis Metrics**: Monitor memory usage, hit rates, and connected clients.
+- **Cluster Nodes**: Realtime view of active app, worker, task, and sandbox nodes through Redis heartbeats.
+- **Task And Workflow Monitoring**: Inspect async tasks and workflow executions, including the node that processed each execution.
+- **Runtime Monitors**: Inspect Redis, event queue, distributed locks, ACL cache, and cache manager state.
+- **Container Orchestrator**: Manage Docker or Kubernetes worker stacks with leader-only write operations.
+- **Worker Packages**: Configure and dispatch apt, npm, and Python package installation across matching node roles.
+- **Plugin Operations**: List installed plugins and force-disable or force-remove broken plugin registry records.
+
+## Architecture Flow
+
+1. `src/index.ts` exports the server package. The client package is exposed through `src/client/index.tsx`.
+2. The client registers the `Cluster Manager` settings page and renders `ClusterManagerLayout`, which groups the admin tools into tabs.
+3. The server `beforeLoad()` imports all collection definitions from `src/server/collections`, including config collections and resource-only collection stubs needed by workflow/ACL lookups.
+4. The server `load()` wires runtime services and APIs: Redis node registry, PubSub adapter, Redis lock adapter polyfill, PubSub subscribers, resource actions, plugin force operations, ACL snippet, ACL cache middleware, health endpoint, and orchestrator initialization.
+5. `afterStart` starts node heartbeats, worker package auto-install, and leader election. `beforeStop` stops the node registry and releases leadership.
+6. Shared runtime state uses Redis and PubSub for heartbeats, execution-node mapping, restart/log/package commands, and package status. Durable settings stay in database collections.
 
 ## Usage
 1. Enable the plugin.
 2. Only accessible to Super Admins.
-3. Navigate to System Settings -> Worker Monitor.
+3. Navigate to System Settings -> Cluster Manager.
 4. Use the dashboard to troubleshoot performance issues, retry failed background tasks, or clear caches.