@camstack/addon-pipeline-orchestrator 1.1.22 → 1.1.23

package/dist/index.js

@@ -24982,6 +24982,15 @@ var OrchestratorDiagnosticsSchema = object({
 	activeDetectionCount: number().int().min(0)
 });
 var pipelineOrchestratorActions = defineCustomActions({ dumpState: customAction(_void(), OrchestratorDiagnosticsSchema) });
+/**
+* Sentinel returned by `buildDetectionConfig` when the profile-slot READ
+* itself failed transiently (e.g. `listAllProfileSlots` discovery timeout
+* while the stream-broker is (re)starting) — as opposed to `null`, which
+* means "genuinely no assigned slot / not configured". Callers MUST treat
+* this differently from `null`: never stop active detection on a transient
+* read failure (the slots almost certainly still exist), and schedule a retry.
+*/
+var TRANSIENT_SLOT_READ = Symbol("transient-slot-read");
 var PipelineOrchestratorAddon = class PipelineOrchestratorAddon extends BaseAddon {
 	/** This node's Moleculer nodeId (from this.ctx.kernel.localNodeId). */
 	localNodeId = "hub";
@@ -25205,6 +25214,15 @@ var PipelineOrchestratorAddon = class PipelineOrchestratorAddon extends BaseAddo
 	loadShedResumeTimer = null;
 	/** Per-camera guard so repeated low-fps snapshots don't stack relocate/pause. */
 	shedInFlight = /* @__PURE__ */ new Set();
+	/**
+	* Per-device backoff for retrying {@link handleDeviceRegistered} after a
+	* TRANSIENT profile-slot read failure ({@link TRANSIENT_SLOT_READ}). Timers
+	* are `unref`'d and cleared on shutdown / device removal.
+	*/
+	slotReadRetryTimers = /* @__PURE__ */ new Map();
+	slotReadRetryAttempts = /* @__PURE__ */ new Map();
+	static SLOT_READ_MAX_RETRIES = 6;
+	static SLOT_READ_RETRY_BASE_MS = 500;
 	/** Pending `scheduleReconcile` debounce timer. */
 	reconcileTimer = null;
 	/** True while `reconcileDispatch` is awaiting the RPC round-trip. */
@@ -25672,6 +25690,9 @@ var PipelineOrchestratorAddon = class PipelineOrchestratorAddon extends BaseAddo
 			clearTimeout(this.pendingRetryDebounceTimer);
 			this.pendingRetryDebounceTimer = null;
 		}
+		for (const t of this.slotReadRetryTimers.values()) clearTimeout(t);
+		this.slotReadRetryTimers.clear();
+		this.slotReadRetryAttempts.clear();
 		this.unsubDeviceRegistered?.();
 		this.unsubDeviceRegistered = null;
 		this.unsubDeviceUnregistered?.();
@@ -26961,6 +26982,47 @@ var PipelineOrchestratorAddon = class PipelineOrchestratorAddon extends BaseAddo
 		}
 	}
 	/**
+	* Bounded per-device retry of {@link handleDeviceRegistered} after a
+	* TRANSIENT profile-slot read failure ({@link TRANSIENT_SLOT_READ}). The
+	* read (`listAllProfileSlots`) can time out during a broker (re)start; we
+	* must neither stop a running camera nor leave a not-yet-started camera
+	* stranded — so re-run the registration handler with exponential backoff
+	* until the read succeeds or the budget is exhausted. Reset on any
+	* successful read ({@link clearSlotReadRetry}).
+	*/
+	scheduleSlotReadRetry(deviceId) {
+		const attempts = this.slotReadRetryAttempts.get(deviceId) ?? 0;
+		if (attempts >= PipelineOrchestratorAddon.SLOT_READ_MAX_RETRIES) {
+			this.ctx.logger.warn("slot-read retry budget exhausted — leaving detection as-is", { tags: { deviceId } });
+			this.slotReadRetryAttempts.delete(deviceId);
+			return;
+		}
+		const existing = this.slotReadRetryTimers.get(deviceId);
+		if (existing) clearTimeout(existing);
+		const delay = PipelineOrchestratorAddon.SLOT_READ_RETRY_BASE_MS * 2 ** attempts;
+		this.slotReadRetryAttempts.set(deviceId, attempts + 1);
+		const timer = setTimeout(() => {
+			this.slotReadRetryTimers.delete(deviceId);
+			this.handleDeviceRegistered(deviceId).catch((err) => {
+				this.ctx.logger.debug("slot-read retry: handleDeviceRegistered failed", {
+					tags: { deviceId },
+					meta: { error: errMsg(err) }
+				});
+			});
+		}, delay);
+		if (typeof timer === "object" && timer !== null && "unref" in timer) timer.unref();
+		this.slotReadRetryTimers.set(deviceId, timer);
+	}
+	/** Cancel any pending slot-read retry + reset the budget for a device. */
+	clearSlotReadRetry(deviceId) {
+		const existing = this.slotReadRetryTimers.get(deviceId);
+		if (existing) {
+			clearTimeout(existing);
+			this.slotReadRetryTimers.delete(deviceId);
+		}
+		this.slotReadRetryAttempts.delete(deviceId);
+	}
+	/**
 	* Coalesce bursts of capacity/eligibility/readiness signals into a single
 	* `retryPendingDispatches` pass. Mirrors `scheduleReconcile`, but on a
 	* dedicated (longer) debounce so a raise-cap / node-connect flurry doesn't
@@ -28837,7 +28899,7 @@ var PipelineOrchestratorAddon = class PipelineOrchestratorAddon extends BaseAddo
 	*/
 	async fetchAssignedProfiles(deviceId) {
 		const api = this.api;
-		if (!api) return /* @__PURE__ */ new Map();
+		if (!api) return null;
 		try {
 			const slots = await api.streamBroker.listAllProfileSlots.query();
 			const out = /* @__PURE__ */ new Map();
@@ -28847,8 +28909,12 @@ var PipelineOrchestratorAddon = class PipelineOrchestratorAddon extends BaseAddo
 				out.set(slot.profile, slot.sourceCamStreamId);
 			}
 			return out;
-		} catch {
-			return /* @__PURE__ */ new Map();
+		} catch (err) {
+			this.ctx.logger.debug("fetchAssignedProfiles: slot read failed (transient) — treating as unknown", {
+				tags: { deviceId },
+				meta: { error: errMsg(err) }
+			});
+			return null;
 		}
 	}
 	/**
@@ -28867,6 +28933,7 @@ var PipelineOrchestratorAddon = class PipelineOrchestratorAddon extends BaseAddo
 	*/
 	async buildDetectionConfig(deviceId) {
 		const assigned = await this.fetchAssignedProfiles(deviceId);
+		if (assigned === null) return TRANSIENT_SLOT_READ;
 		if (assigned.size === 0) return null;
 		const resolved = await this.resolveDeviceDetectionSettings(deviceId);
 		if (!resolved) return null;
@@ -29010,6 +29077,7 @@ var PipelineOrchestratorAddon = class PipelineOrchestratorAddon extends BaseAddo
 	}
 	/** Stop detection and purge all persisted config for a removed device. */
 	async handleDeviceUnregistered(deviceId) {
+		this.clearSlotReadRetry(deviceId);
 		await this.stopDetection(deviceId);
 		this.zonesProvider?.forgetDevice(deviceId);
 		this.zoneRulesProvider?.forgetDevice(deviceId);
@@ -29031,6 +29099,12 @@ var PipelineOrchestratorAddon = class PipelineOrchestratorAddon extends BaseAddo
 		const log = this.ctx.logger.withTags({ deviceId });
 		log.info("handleDeviceRegistered", { meta: { phase: "handleDeviceRegistered" } });
 		const config = await this.buildDetectionConfig(deviceId);
+		if (config === TRANSIENT_SLOT_READ) {
+			log.warn("Profile-slot read failed transiently — keeping detection, scheduling retry");
+			this.scheduleSlotReadRetry(deviceId);
+			return;
+		}
+		this.clearSlotReadRetry(deviceId);
 		log.info("[pipeline-orchestrator] buildDetectionConfig", config ? { meta: {
 			enabled: config.enabled,
 			motionStreamId: config.motionStreamId,
@@ -29096,6 +29170,12 @@ var PipelineOrchestratorAddon = class PipelineOrchestratorAddon extends BaseAddo
 		const log = this.ctx.logger.withTags({ deviceId });
 		if (this.activeDetections.has(deviceId)) await this.stopDetection(deviceId);
 		const config = await this.buildDetectionConfig(deviceId);
+		if (config === TRANSIENT_SLOT_READ) {
+			log.warn("Settings changed — slot read failed transiently, scheduling retry");
+			this.scheduleSlotReadRetry(deviceId);
+			return;
+		}
+		this.clearSlotReadRetry(deviceId);
 		if (!config) {
 			log.info("Settings changed — no assigned slot, detection stopped");
 			return;

package/dist/index.mjs

@@ -24978,6 +24978,15 @@ var OrchestratorDiagnosticsSchema = object({
 	activeDetectionCount: number().int().min(0)
 });
 var pipelineOrchestratorActions = defineCustomActions({ dumpState: customAction(_void(), OrchestratorDiagnosticsSchema) });
+/**
+* Sentinel returned by `buildDetectionConfig` when the profile-slot READ
+* itself failed transiently (e.g. `listAllProfileSlots` discovery timeout
+* while the stream-broker is (re)starting) — as opposed to `null`, which
+* means "genuinely no assigned slot / not configured". Callers MUST treat
+* this differently from `null`: never stop active detection on a transient
+* read failure (the slots almost certainly still exist), and schedule a retry.
+*/
+var TRANSIENT_SLOT_READ = Symbol("transient-slot-read");
 var PipelineOrchestratorAddon = class PipelineOrchestratorAddon extends BaseAddon {
 	/** This node's Moleculer nodeId (from this.ctx.kernel.localNodeId). */
 	localNodeId = "hub";
@@ -25201,6 +25210,15 @@ var PipelineOrchestratorAddon = class PipelineOrchestratorAddon extends BaseAddo
 	loadShedResumeTimer = null;
 	/** Per-camera guard so repeated low-fps snapshots don't stack relocate/pause. */
 	shedInFlight = /* @__PURE__ */ new Set();
+	/**
+	* Per-device backoff for retrying {@link handleDeviceRegistered} after a
+	* TRANSIENT profile-slot read failure ({@link TRANSIENT_SLOT_READ}). Timers
+	* are `unref`'d and cleared on shutdown / device removal.
+	*/
+	slotReadRetryTimers = /* @__PURE__ */ new Map();
+	slotReadRetryAttempts = /* @__PURE__ */ new Map();
+	static SLOT_READ_MAX_RETRIES = 6;
+	static SLOT_READ_RETRY_BASE_MS = 500;
 	/** Pending `scheduleReconcile` debounce timer. */
 	reconcileTimer = null;
 	/** True while `reconcileDispatch` is awaiting the RPC round-trip. */
@@ -25668,6 +25686,9 @@ var PipelineOrchestratorAddon = class PipelineOrchestratorAddon extends BaseAddo
 			clearTimeout(this.pendingRetryDebounceTimer);
 			this.pendingRetryDebounceTimer = null;
 		}
+		for (const t of this.slotReadRetryTimers.values()) clearTimeout(t);
+		this.slotReadRetryTimers.clear();
+		this.slotReadRetryAttempts.clear();
 		this.unsubDeviceRegistered?.();
 		this.unsubDeviceRegistered = null;
 		this.unsubDeviceUnregistered?.();
@@ -26957,6 +26978,47 @@ var PipelineOrchestratorAddon = class PipelineOrchestratorAddon extends BaseAddo
 		}
 	}
 	/**
+	* Bounded per-device retry of {@link handleDeviceRegistered} after a
+	* TRANSIENT profile-slot read failure ({@link TRANSIENT_SLOT_READ}). The
+	* read (`listAllProfileSlots`) can time out during a broker (re)start; we
+	* must neither stop a running camera nor leave a not-yet-started camera
+	* stranded — so re-run the registration handler with exponential backoff
+	* until the read succeeds or the budget is exhausted. Reset on any
+	* successful read ({@link clearSlotReadRetry}).
+	*/
+	scheduleSlotReadRetry(deviceId) {
+		const attempts = this.slotReadRetryAttempts.get(deviceId) ?? 0;
+		if (attempts >= PipelineOrchestratorAddon.SLOT_READ_MAX_RETRIES) {
+			this.ctx.logger.warn("slot-read retry budget exhausted — leaving detection as-is", { tags: { deviceId } });
+			this.slotReadRetryAttempts.delete(deviceId);
+			return;
+		}
+		const existing = this.slotReadRetryTimers.get(deviceId);
+		if (existing) clearTimeout(existing);
+		const delay = PipelineOrchestratorAddon.SLOT_READ_RETRY_BASE_MS * 2 ** attempts;
+		this.slotReadRetryAttempts.set(deviceId, attempts + 1);
+		const timer = setTimeout(() => {
+			this.slotReadRetryTimers.delete(deviceId);
+			this.handleDeviceRegistered(deviceId).catch((err) => {
+				this.ctx.logger.debug("slot-read retry: handleDeviceRegistered failed", {
+					tags: { deviceId },
+					meta: { error: errMsg(err) }
+				});
+			});
+		}, delay);
+		if (typeof timer === "object" && timer !== null && "unref" in timer) timer.unref();
+		this.slotReadRetryTimers.set(deviceId, timer);
+	}
+	/** Cancel any pending slot-read retry + reset the budget for a device. */
+	clearSlotReadRetry(deviceId) {
+		const existing = this.slotReadRetryTimers.get(deviceId);
+		if (existing) {
+			clearTimeout(existing);
+			this.slotReadRetryTimers.delete(deviceId);
+		}
+		this.slotReadRetryAttempts.delete(deviceId);
+	}
+	/**
 	* Coalesce bursts of capacity/eligibility/readiness signals into a single
 	* `retryPendingDispatches` pass. Mirrors `scheduleReconcile`, but on a
 	* dedicated (longer) debounce so a raise-cap / node-connect flurry doesn't
@@ -28833,7 +28895,7 @@ var PipelineOrchestratorAddon = class PipelineOrchestratorAddon extends BaseAddo
 	*/
 	async fetchAssignedProfiles(deviceId) {
 		const api = this.api;
-		if (!api) return /* @__PURE__ */ new Map();
+		if (!api) return null;
 		try {
 			const slots = await api.streamBroker.listAllProfileSlots.query();
 			const out = /* @__PURE__ */ new Map();
@@ -28843,8 +28905,12 @@ var PipelineOrchestratorAddon = class PipelineOrchestratorAddon extends BaseAddo
 				out.set(slot.profile, slot.sourceCamStreamId);
 			}
 			return out;
-		} catch {
-			return /* @__PURE__ */ new Map();
+		} catch (err) {
+			this.ctx.logger.debug("fetchAssignedProfiles: slot read failed (transient) — treating as unknown", {
+				tags: { deviceId },
+				meta: { error: errMsg(err) }
+			});
+			return null;
 		}
 	}
 	/**
@@ -28863,6 +28929,7 @@ var PipelineOrchestratorAddon = class PipelineOrchestratorAddon extends BaseAddo
 	*/
 	async buildDetectionConfig(deviceId) {
 		const assigned = await this.fetchAssignedProfiles(deviceId);
+		if (assigned === null) return TRANSIENT_SLOT_READ;
 		if (assigned.size === 0) return null;
 		const resolved = await this.resolveDeviceDetectionSettings(deviceId);
 		if (!resolved) return null;
@@ -29006,6 +29073,7 @@ var PipelineOrchestratorAddon = class PipelineOrchestratorAddon extends BaseAddo
 	}
 	/** Stop detection and purge all persisted config for a removed device. */
 	async handleDeviceUnregistered(deviceId) {
+		this.clearSlotReadRetry(deviceId);
 		await this.stopDetection(deviceId);
 		this.zonesProvider?.forgetDevice(deviceId);
 		this.zoneRulesProvider?.forgetDevice(deviceId);
@@ -29027,6 +29095,12 @@ var PipelineOrchestratorAddon = class PipelineOrchestratorAddon extends BaseAddo
 		const log = this.ctx.logger.withTags({ deviceId });
 		log.info("handleDeviceRegistered", { meta: { phase: "handleDeviceRegistered" } });
 		const config = await this.buildDetectionConfig(deviceId);
+		if (config === TRANSIENT_SLOT_READ) {
+			log.warn("Profile-slot read failed transiently — keeping detection, scheduling retry");
+			this.scheduleSlotReadRetry(deviceId);
+			return;
+		}
+		this.clearSlotReadRetry(deviceId);
 		log.info("[pipeline-orchestrator] buildDetectionConfig", config ? { meta: {
 			enabled: config.enabled,
 			motionStreamId: config.motionStreamId,
@@ -29092,6 +29166,12 @@ var PipelineOrchestratorAddon = class PipelineOrchestratorAddon extends BaseAddo
 		const log = this.ctx.logger.withTags({ deviceId });
 		if (this.activeDetections.has(deviceId)) await this.stopDetection(deviceId);
 		const config = await this.buildDetectionConfig(deviceId);
+		if (config === TRANSIENT_SLOT_READ) {
+			log.warn("Settings changed — slot read failed transiently, scheduling retry");
+			this.scheduleSlotReadRetry(deviceId);
+			return;
+		}
+		this.clearSlotReadRetry(deviceId);
 		if (!config) {
 			log.info("Settings changed — no assigned slot, detection stopped");
 			return;

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@camstack/addon-pipeline-orchestrator",
-  "version": "1.1.22",
+  "version": "1.1.23",
   "description": "Hub-side camera-to-agent load balancer — tracks runner capacity and dispatches attachCamera calls to the optimal pipeline-runner instance",
   "keywords": [
     "camstack",