Struct AppState 

pub struct AppState {

    pub cluster_config: ClusterConfig,
    pub container_runtime: Arc<dyn Runtime>,
    pub wasm_runtime: Option<Arc<dyn Runtime>>,
    pub services: RwLock<HashMap<String, ServiceState>>,
    pub route_table: SharedRouteTable,
    pub wasm_triggers: SharedWasmTriggers,
    pub registered_nodes: RwLock<HashMap<u64, RegisteredNode>>,
    pub webhooks: WebhookStore,
    pub api_tokens: Vec<String>,
    pub pending_commands: RwLock<HashMap<u64, Vec<Value>>>,
    pub deploy_history: RwLock<DeployHistory>,
    pub acme_manager: Option<AcmeManager>,
    pub cert_resolver: Option<SharedCertResolver>,
    pub container_stats: RwLock<HashMap<String, ContainerStats>>,
    pub store: Option<Arc<ClusterStore>>,
    pub ws_agents: RwLock<HashMap<u64, AgentSender>>,
    pub log_listeners: RwLock<HashMap<String, Sender<(String, bool)>>>,
    pub backup_listeners: RwLock<HashMap<String, Sender<BackupStatusReportData>>>,
    pub network_listeners: RwLock<HashMap<String, Sender<NetworkStatusReportData>>>,
    pub last_backup_results: RwLock<HashMap<u64, LastBackupResult>>,
    pub master_last_backup_result: RwLock<Option<LastBackupResult>>,
    pub webhook_invocations: RwLock<HashMap<String, VecDeque<WebhookInvocation>>>,
    pub exec_sessions: RwLock<HashMap<String, Sender<Vec<u8>>>>,
    pub pending_deploys: RwLock<HashMap<String, Sender<Result<(), String>>>>,
    pub alerts: Option<SharedAlertEngine>,
    pub instance_events: RwLock<HashMap<String, InstanceEventLog>>,
}

Shared state for the control plane, accessible by the API server and reconciler.

Fields

cluster_config: ClusterConfig

Cluster configuration.

container_runtime: Arc<dyn Runtime>

Container runtime (Docker).

wasm_runtime: Option<Arc<dyn Runtime>>

Wasm runtime (wasmtime). Trait object to avoid coupling to concrete type.

services: RwLock<HashMap<String, ServiceState>>

Current service state, keyed by service name.

route_table: SharedRouteTable

Routing table for container workloads, shared with the reverse proxy.

wasm_triggers: SharedWasmTriggers

Wasm HTTP triggers, shared with the reverse proxy.

registered_nodes: RwLock<HashMap<u64, RegisteredNode>>

Registered cluster nodes (M2 in-memory, will move to Raft store).

webhooks: WebhookStore

Webhook configurations for push-triggered deploys.

api_tokens: Vec<String>

API bearer tokens for authentication (empty = allow all).

pending_commands: RwLock<HashMap<u64, Vec<Value>>>

Pending commands for agent nodes, keyed by node_id. Uses serde_json::Value to avoid circular dependency on orca-agent types.

deploy_history: RwLock<DeployHistory>

Deploy history for rollback support.

acme_manager: Option<AcmeManager>

ACME manager for hot cert provisioning (None if no TLS).

cert_resolver: Option<SharedCertResolver>

Dynamic cert resolver shared with the HTTPS listener.

container_stats: RwLock<HashMap<String, ContainerStats>>

Cached container stats, keyed by service name.

store: Option<Arc<ClusterStore>>

Persistent cluster store (redb). None in tests without persistence.

ws_agents: RwLock<HashMap<u64, AgentSender>>

WebSocket senders for connected agent nodes, keyed by node_id.

log_listeners: RwLock<HashMap<String, Sender<(String, bool)>>>

Log stream listeners: request_id → (data, done) sender.

backup_listeners: RwLock<HashMap<String, Sender<BackupStatusReportData>>>

Backup status listeners: request_id → report sender. Used by the /api/v1/cluster/backups handler to collect reports dispatched in parallel to every connected agent.

network_listeners: RwLock<HashMap<String, Sender<NetworkStatusReportData>>>

Network status listeners: request_id → report sender. Same pattern as backup_listeners, used by the /api/v1/cluster/networks handler.

last_backup_results: RwLock<HashMap<u64, LastBackupResult>>

Last completed BackupResult per agent node, recorded as the result arrives over WS. Surfaced alongside the snapshot listing so the dashboard can show a node’s last-failure message without having to scrape logs. Master has its own field — its backups are subprocess- driven, not WS-dispatched.

master_last_backup_result: RwLock<Option<LastBackupResult>>

Last completed master backup result, recorded when run_master_backup finishes. Separate from last_backup_results because the master has no node_id and runs its backups via subprocess rather than WS.

webhook_invocations: RwLock<HashMap<String, VecDeque<WebhookInvocation>>>

Recent webhook invocations, keyed by service_name. Bounded ring buffer of the last 10 deliveries per webhook so the TUI can render a history view without scraping logs. Lost on restart by design — this is operator-visible recent activity, not durable audit.

exec_sessions: RwLock<HashMap<String, Sender<Vec<u8>>>>

Active exec sessions: session_id → output bytes sender (agent → CLI WS).

pending_deploys: RwLock<HashMap<String, Sender<Result<(), String>>>>

Pending deploy result waiters: service_name → oneshot sender. Inserted by queue_remote_deploy, resolved by ws_handler on DeployResult.

alerts: Option<SharedAlertEngine>

AI conversational-alert engine. None when [ai] is not configured. AiMonitor::run (spawned in lib::run_server_with_acme) feeds it open_alert calls; the HTTP /api/v1/alerts/... handlers mutate it in response to operator actions.

instance_events: RwLock<HashMap<String, InstanceEventLog>>

Per-service log of restart triggers and instance failures. Read by the AI alert monitor to populate restart_count_24h and error_count_1h in the cluster context. In-memory only; restart wipes the log (the AI monitor reconstructs alerts as conditions reappear, so a brief gap after restart is acceptable).

Implementations

impl AppState

pub fn new( cluster_config: ClusterConfig, container_runtime: Arc<dyn Runtime>, wasm_runtime: Option<Arc<dyn Runtime>>, route_table: SharedRouteTable, wasm_triggers: SharedWasmTriggers, ) -> Self

Create with shared route table and Wasm triggers (for sharing with the proxy).

pub async fn record_instance_restart(&self, service: &str)

Append a restart event for service and prune old entries. Cheap (one now() + a few VecDeque ops); safe to call from hot paths.

pub async fn record_instance_failure(&self, service: &str)

Append a failure event for service (non-zero exit or health-check failure) and prune old entries.

pub fn with_store(self, store: Arc<ClusterStore>) -> Self

Set persistent store for service state.

pub fn with_alerts(self, engine: SharedAlertEngine) -> Self

Attach the AI alert engine. Builder-style so startup code can wire it only when [ai] is configured.

pub fn with_acme( self, manager: AcmeManager, resolver: SharedCertResolver, ) -> Self

Set ACME manager and cert resolver for hot cert provisioning.