Fix stale data after HA outage recovery

Two bugs caused the module to show outdated data permanently after HA
came back online:

1. refreshTimer was only checked at section level, but config sets it at
   the module config level. The setInterval never started, so there was
   no periodic re-render fallback when the WebSocket died.

2. _closeCircuit replayed queued templates but never reconnected the
   WebSocket. Without WS, no state_changed events fire, so the only
   render path was the (broken) refreshTimer.

Also fixes a race condition in _healthCheck where breaker.state was
briefly set to 'closed' before calling _openCircuit on failure. Now
uses 'half-open' state instead.

MMM-HomeAssistantDisplay.js

@@ -52,15 +52,29 @@ Module.register("MMM-HomeAssistantDisplay", {
 						entity: section.triggerEntities[entity]
 					});
 				}
-				// Set up a timer to trigger re-rendering outside of any entity state update
-				if (section.refreshTimer) {
-					setInterval(()=> {
-						this.renderTemplates("timeout");
-						this.updateDom();
-					}, section.refreshTimer * 1000);
+			}
+		}
+
+		// Refresh timer: check section-level first, then fall back to config-level.
+		// One interval per module instance — renderTemplates already hits all sections.
+		var refreshInterval = null;
+		if (this.config.sections) {
+			for (const sectioid in this.config.sections) {
+				if (this.config.sections[sectioid].refreshTimer) {
+					refreshInterval = this.config.sections[sectioid].refreshTimer;
+					break;
 				}
 			}
 		}
+		if (!refreshInterval && this.config.refreshTimer) {
+			refreshInterval = this.config.refreshTimer;
+		}
+		if (refreshInterval) {
+			setInterval(() => {
+				this.renderTemplates("refreshTimer");
+				this.updateDom();
+			}, refreshInterval * 1000);
+		}
 		this.renderTemplates("foo");
 		self.updateDom(self.config.animationSpeed);
 	},

node_helper.js

@@ -3,6 +3,12 @@
  *
  * By Brian Towles
  * MIT Licensed.
+ *
+ * Resilience improvements by James (2026-02-27):
+ * - Circuit breaker: stops all template evals when HA is unreachable
+ * - Exponential backoff on retries (no more 30 simultaneous retry timers)
+ * - Coalesced health checks instead of per-section retries
+ * - Graceful degradation: sections show stale data instead of freezing
  */
 var backoff = require('backoff')
 const NodeHelper = require("node_helper");
@@ -24,6 +30,10 @@ module.exports = NodeHelper.create({
 	evaluateTemplate,
 	onWebsocketCloseEvent,
 	backoffWSConnection,
+	_getBreaker,
+	_openCircuit,
+	_closeCircuit,
+	_healthCheck,
 });
 
 function start() {
@@ -32,6 +42,10 @@ function start() {
 		this.logger.debug("MMM-HomeAssistantDisplay helper started...");
 	}
 	this.connections = {};
+
+	// Circuit breaker state: tracks HA reachability per connection
+	// When open, template evaluations are skipped entirely (no requests, no error spam)
+	this._circuitBreaker = {};  // identifier -> { state: 'closed'|'open', failCount: 0, nextRetryAt: 0, retryTimer: null }
 }
 
 function stop() {
@@ -67,10 +81,11 @@ function socketNotificationReceived(notification, payload) {
 			this.evaluateTemplate(payload).then((ret) => {
 				this.sendSocketNotification("MODULE_DISPLAY_RENDERED", ret);
 			}).catch((err) => {
-				this.logger.error(
-					"Unable to evaluate template",
-					err
-				);
+				// Only log if circuit is closed (avoid spam when HA is down)
+				const cb = this._getBreaker(payload.identifier);
+				if (cb.state === 'closed') {
+					this.logger.error("Unable to evaluate template", err);
+				}
 			});
 			break;
 		case "RENDER_SECTION_DISPLAY_TEMPLATE":
@@ -80,24 +95,172 @@ function socketNotificationReceived(notification, payload) {
 					section: payload.section
 				});
 			}).catch((err) => {
-				this.logger.error(
-					"unable to evaluate section template",
-					err
-				);
+				const cb = this._getBreaker(payload.identifier);
+				if (cb.state === 'closed') {
+					this.logger.error("unable to evaluate section template", err);
+				}
 			});
 			break;			
 	}
 }
 
+// Circuit breaker helpers
+function _getBreaker(identifier) {
+	if (!this._circuitBreaker[identifier]) {
+		this._circuitBreaker[identifier] = {
+			state: 'closed',      // closed = healthy, open = HA unreachable
+			failCount: 0,
+			nextRetryAt: 0,
+			retryTimer: null,
+			pendingQueue: [],     // queued payloads to retry when circuit closes
+		};
+	}
+	return this._circuitBreaker[identifier];
+}
+
+function _openCircuit(identifier) {
+	const breaker = this._getBreaker(identifier);
+	if (breaker.state === 'open') return; // already open
+	// Accepts 'closed' (first failure) or 'half-open' (health check retry failed)
+
+	breaker.state = 'open';
+	breaker.failCount++;
+	// Exponential backoff: 15s, 30s, 60s, 120s, max 300s
+	const delay = Math.min(15000 * Math.pow(2, breaker.failCount - 1), 300000);
+	breaker.nextRetryAt = Date.now() + delay;
+
+	this.logger.info(`Circuit OPEN for ${identifier} — HA unreachable. Next health check in ${delay / 1000}s`);
+
+	// Single coalesced health check timer (not per-section!)
+	if (breaker.retryTimer) clearTimeout(breaker.retryTimer);
+	breaker.retryTimer = setTimeout(() => {
+		this._healthCheck(identifier);
+	}, delay);
+}
+
+function _closeCircuit(identifier) {
+	const breaker = this._getBreaker(identifier);
+	if (breaker.state === 'closed') return;
+
+	this.logger.info(`Circuit CLOSED for ${identifier} — HA is reachable again`);
+	breaker.state = 'closed';
+	breaker.failCount = 0;
+	breaker.nextRetryAt = 0;
+	if (breaker.retryTimer) {
+		clearTimeout(breaker.retryTimer);
+		breaker.retryTimer = null;
+	}
+
+	// Reconnect WebSocket — it likely died during the outage and won't
+	// recover on its own because the frontend only sends RECONNECT_WS in
+	// response to HASSWS_DISCONNECTED, which the node_helper may never
+	// have emitted if the socket died silently.
+	const conn = this.connections[identifier];
+	if (conn && conn.connectionConfig) {
+		this.logger.info(`Reconnecting WebSocket for ${identifier} after circuit recovery`);
+		this.backoffWSConnection(identifier, conn.connectionConfig);
+	}
+
+	// Replay queued template evaluations
+	const queue = breaker.pendingQueue.splice(0);
+	if (queue.length > 0) {
+		this.logger.info(`Replaying ${queue.length} queued template evaluations for ${identifier}`);
+		for (const item of queue) {
+			this.evaluateTemplate(item.payload).then((ret) => {
+				if (item.payload.section !== undefined) {
+					this.sendSocketNotification("SECTION_DISPLAY_RENDERED", {
+						...ret,
+						section: item.payload.section
+					});
+				} else {
+					this.sendSocketNotification("MODULE_DISPLAY_RENDERED", ret);
+				}
+			}).catch(() => {
+				// Circuit will re-open if this fails
+			});
+		}
+	}
+}
+
+async function _healthCheck(identifier) {
+	const breaker = this._getBreaker(identifier);
+	const hass = this.connections[identifier] && this.connections[identifier].hass;
+	if (!hass) return;
+
+	this.logger.info(`Health check for ${identifier}...`);
+	try {
+		// Simple template render as health probe
+		await Promise.race([
+			hass.templates.render("{{ 1 }}"),
+			new Promise((_, reject) =>
+				setTimeout(() => reject(new Error('Health check timeout')), 10000)
+			)
+		]);
+		// Success — close the circuit
+		this._closeCircuit(identifier);
+	} catch (err) {
+		this.logger.info(`Health check failed for ${identifier}: ${err.message}`);
+		// Re-arm with increased backoff — reset state so _openCircuit can fire
+		breaker.state = 'half-open';
+		this._openCircuit(identifier);
+	}
+}
+
 async function evaluateTemplate(payload) {
 	if (config.debuglogging) {
 		this.logger.debug(`Evaluating template for ${payload.template}`);
 	}
+
+	// Circuit breaker check: if HA is known-unreachable, skip entirely
+	const breaker = this._getBreaker(payload.identifier);
+	if (breaker.state === 'open') {
+		// Queue this request for replay when circuit closes (deduplicated by section)
+		const isDuplicate = breaker.pendingQueue.some(
+			(item) => item.payload.section === payload.section && item.payload.template === payload.template
+		);
+		if (!isDuplicate) {
+			breaker.pendingQueue.push({ payload });
+			// Cap queue size to prevent memory growth
+			if (breaker.pendingQueue.length > 20) {
+				breaker.pendingQueue.shift();
+			}
+		}
+		throw new Error('Circuit open — HA unreachable, skipping template evaluation');
+	}
+
 	const hass = this.connections[payload.identifier].hass;
-	const response = await hass.templates.render(payload.template);
-	return {
-		identifier: payload.identifier,
-		render: response
+
+	try {
+		// Wrap template call with timeout
+		const response = await Promise.race([
+			hass.templates.render(payload.template),
+			new Promise((_, reject) =>
+				setTimeout(() => reject(new Error('Template evaluation timeout')), 10000)
+			)
+		]);
+
+		// Success — ensure circuit is closed (resets fail count)
+		if (breaker.failCount > 0) {
+			this._closeCircuit(payload.identifier);
+		}
+
+		return {
+			identifier: payload.identifier,
+			render: response
+		};
+	} catch (err) {
+		// Check if this is a connectivity error (not a template syntax error)
+		const isConnectivityError = /ENETUNREACH|ECONNREFUSED|ECONNRESET|ETIMEDOUT|EHOSTUNREACH|timeout/i.test(err.message);
+
+		if (isConnectivityError) {
+			this.logger.error(`Template evaluation failed (connectivity): ${err.message}`);
+			this._openCircuit(payload.identifier);
+		} else {
+			// Template/logic error — log but don't trip the breaker
+			this.logger.error(`Template evaluation failed: ${err.message}`);
+		}
+
+		throw err;
 	}
 }
 
@@ -125,7 +288,8 @@ async function connect(payload) {
 	this.logger.info(`HomeAssistant connected for ${payload.identifier}`);
 	this.connections[payload.identifier] = {
 		hass,
-		entities: []
+		entities: [],
+		connectionConfig,
 	};
 
 	await this.backoffWSConnection(payload.identifier, connectionConfig)
@@ -192,12 +356,8 @@ async function reconnectWebsocket(payload) {
 }
 
 function onStateChangedEvent(event) {
-	if (config.debuglogging) {
-		this.logger.debug(`Got state change for ${event.data.entity_id}`);
-	}
 	for (const connection in this.connections) {
 		if (this.connections[connection].entities.includes(event.data.entity_id)) {
-			this.logger.debug(`Found listening connection (${connection}) for entity ${event.data.entity_id}`);
 			this.sendSocketNotification("CHANGED_STATE", {
 				identifier: connection,
 				cause: event.data.entity_id,
@@ -209,7 +369,7 @@ function onStateChangedEvent(event) {
 function onWebsocketCloseEvent(event) {
 	for (const connection in this.connections) {
 		if (event.target == this.connections[connection].websocket.rawClient.ws) {
-			this.logger.debug(`Hass WS Disconnected (${connection})`);
+			this.logger.info(`Hass WS Disconnected (${connection})`);
 			this.sendSocketNotification("HASSWS_DISCONNECTED", {
 				identifier: connection,
 			});