@kici-dev/engine 0.12.0 → 0.14.0

package/dist/protocol/messages/peer.d.ts

@@ -131,6 +131,15 @@ export declare const peerHeartbeatSchema: z.ZodObject<{
     runningAsUid: z.ZodOptional<z.ZodNullable<z.ZodNumber>>;
     version: z.ZodOptional<z.ZodString>;
 }, z.core.$strip>;
+/**
+ * The spawn-retry budget a worker applies to one rerouted job: how many agent
+ * spawns it attempts, and how long it waits after a failed one. The sending
+ * coordinator resolves it per org and enforces the same budget on its side.
+ */
+export declare const rerouteSpawnRetrySchema: z.ZodObject<{
+    maxAttempts: z.ZodNumber;
+    backoffMs: z.ZodNumber;
+}, z.core.$strip>;
 /** Request to reroute a job to another orchestrator (no local agent can handle it). */
 export declare const jobRerouteSchema: z.ZodObject<{
     type: z.ZodLiteral<"job.reroute">;
@@ -164,6 +173,10 @@ export declare const jobRerouteSchema: z.ZodObject<{
     }, z.core.$strip>], "kind">>>;
     triedConnections: z.ZodArray<z.ZodString>;
     maxHops: z.ZodNumber;
+    spawnRetry: z.ZodOptional<z.ZodObject<{
+        maxAttempts: z.ZodNumber;
+        backoffMs: z.ZodNumber;
+    }, z.core.$strip>>;
     coordinatorId: z.ZodString;
     requestId: z.ZodOptional<z.ZodString>;
     traceId: z.ZodOptional<z.ZodString>;
@@ -351,8 +364,177 @@ export declare const peerConfigReloadResponseSchema: z.ZodObject<{
     restartRequired: z.ZodOptional<z.ZodArray<z.ZodString>>;
     fieldsChanged: z.ZodOptional<z.ZodArray<z.ZodString>>;
 }, z.core.$strip>;
+/** How a live Firecracker VM relates to the orchestrator on its host. */
+export declare const ScalerVmStatus: z.ZodEnum<{
+    orphaned: "orphaned";
+    tracked: "tracked";
+    unverified: "unverified";
+}>;
+export type ScalerVmStatus = z.infer<typeof ScalerVmStatus>;
+/** What tracks a live VM. `bound-job` is answered from a coordinator's dispatch queue. */
+export declare const ScalerVmTracker: z.ZodEnum<{
+    backend: "backend";
+    "bound-job": "bound-job";
+    registered: "registered";
+    spawning: "spawning";
+}>;
+export type ScalerVmTracker = z.infer<typeof ScalerVmTracker>;
+/** What a stop did to one VM. Only `stopped` signalled anything. */
+export declare const ScalerVmStopOutcome: z.ZodEnum<{
+    error: "error";
+    "not-found": "not-found";
+    "not-live": "not-live";
+    stopped: "stopped";
+    tracked: "tracked";
+    unverified: "unverified";
+}>;
+export type ScalerVmStopOutcome = z.infer<typeof ScalerVmStopOutcome>;
+/** What a scaler orphan request asks the target node to do. */
+export declare const ScalerOrphansAction: z.ZodEnum<{
+    list: "list";
+    stop: "stop";
+}>;
+export type ScalerOrphansAction = z.infer<typeof ScalerOrphansAction>;
+/** One live Firecracker VM on a node, as its orchestrator sees it. */
+export declare const scalerLiveVmSchema: z.ZodObject<{
+    vmId: z.ZodString;
+    scaler: z.ZodString;
+    pid: z.ZodNumber;
+    startedAt: z.ZodString;
+    ageSeconds: z.ZodNumber;
+    chrootDir: z.ZodString;
+    status: z.ZodEnum<{
+        orphaned: "orphaned";
+        tracked: "tracked";
+        unverified: "unverified";
+    }>;
+    trackedBy: z.ZodArray<z.ZodEnum<{
+        backend: "backend";
+        "bound-job": "bound-job";
+        registered: "registered";
+        spawning: "spawning";
+    }>>;
+    reason: z.ZodString;
+}, z.core.$strip>;
+export type ScalerLiveVm = z.infer<typeof scalerLiveVmSchema>;
+/** The result of stopping one VM. `pid` is the process that was (or would be) signalled. */
+export declare const scalerVmStopResultSchema: z.ZodObject<{
+    vmId: z.ZodString;
+    outcome: z.ZodEnum<{
+        error: "error";
+        "not-found": "not-found";
+        "not-live": "not-live";
+        stopped: "stopped";
+        tracked: "tracked";
+        unverified: "unverified";
+    }>;
+    pid: z.ZodOptional<z.ZodNumber>;
+    detail: z.ZodString;
+}, z.core.$strip>;
+export type ScalerVmStopResult = z.infer<typeof scalerVmStopResultSchema>;
+/**
+ * Scaler orphan request: forwarded by a coordinator to the node an operator
+ * targeted with `kici-admin scaler orphans --target`. The node answers from
+ * its own host and its own tracking with peer.scaler.orphans.response.
+ */
+export declare const peerScalerOrphansRequestSchema: z.ZodObject<{
+    type: z.ZodLiteral<"peer.scaler.orphans.request">;
+    messageId: z.ZodString;
+    action: z.ZodEnum<{
+        list: "list";
+        stop: "stop";
+    }>;
+    vmIds: z.ZodOptional<z.ZodArray<z.ZodString>>;
+}, z.core.$strip>;
+/**
+ * Scaler orphan response. `vms` answers a `list`, `results` answers a `stop`;
+ * `ok: false` carries the node's `error` instead.
+ */
+export declare const peerScalerOrphansResponseSchema: z.ZodObject<{
+    type: z.ZodLiteral<"peer.scaler.orphans.response">;
+    messageId: z.ZodString;
+    ok: z.ZodBoolean;
+    error: z.ZodOptional<z.ZodString>;
+    firecrackerScalers: z.ZodOptional<z.ZodArray<z.ZodString>>;
+    vms: z.ZodOptional<z.ZodArray<z.ZodObject<{
+        vmId: z.ZodString;
+        scaler: z.ZodString;
+        pid: z.ZodNumber;
+        startedAt: z.ZodString;
+        ageSeconds: z.ZodNumber;
+        chrootDir: z.ZodString;
+        status: z.ZodEnum<{
+            orphaned: "orphaned";
+            tracked: "tracked";
+            unverified: "unverified";
+        }>;
+        trackedBy: z.ZodArray<z.ZodEnum<{
+            backend: "backend";
+            "bound-job": "bound-job";
+            registered: "registered";
+            spawning: "spawning";
+        }>>;
+        reason: z.ZodString;
+    }, z.core.$strip>>>;
+    results: z.ZodOptional<z.ZodArray<z.ZodObject<{
+        vmId: z.ZodString;
+        outcome: z.ZodEnum<{
+            error: "error";
+            "not-found": "not-found";
+            "not-live": "not-live";
+            stopped: "stopped";
+            tracked: "tracked";
+            unverified: "unverified";
+        }>;
+        pid: z.ZodOptional<z.ZodNumber>;
+        detail: z.ZodString;
+    }, z.core.$strip>>>;
+}, z.core.$strip>;
+/**
+ * What forgetting a departed peer did on one coordinator. `recent`: the peer
+ * was heard from inside the window this coordinator still treats it as alive
+ * for (the longer of the stale window and the reroute flap grace, or the
+ * backstop grace of a peer that adopted provisions), so it may be partitioned
+ * rather than gone. `acknowledgement-required`: forgetting it would switch this
+ * coordinator's event-provision backstop back on, and the request did not
+ * acknowledge that. Every outcome but `forgotten` keeps the peer.
+ */
+export declare const PeerForgetOutcome: z.ZodEnum<{
+    "acknowledgement-required": "acknowledgement-required";
+    connected: "connected";
+    error: "error";
+    forgotten: "forgotten";
+    "not-found": "not-found";
+    recent: "recent";
+}>;
+export type PeerForgetOutcome = z.infer<typeof PeerForgetOutcome>;
+/**
+ * Forget request: a coordinator that forgot a departed peer (`kici-admin peer
+ * forget`) asks each connected sibling coordinator to drop it from its own
+ * live peer registry too.
+ */
+export declare const peerForgetRequestSchema: z.ZodObject<{
+    type: z.ZodLiteral<"peer.forget.request">;
+    messageId: z.ZodString;
+    instanceId: z.ZodString;
+    acknowledgeBackstop: z.ZodOptional<z.ZodBoolean>;
+}, z.core.$strip>;
+/** Forget response: what the sibling did with the request. */
+export declare const peerForgetResponseSchema: z.ZodObject<{
+    type: z.ZodLiteral<"peer.forget.response">;
+    messageId: z.ZodString;
+    outcome: z.ZodEnum<{
+        "acknowledgement-required": "acknowledgement-required";
+        connected: "connected";
+        error: "error";
+        forgotten: "forgotten";
+        "not-found": "not-found";
+        recent: "recent";
+    }>;
+    detail: z.ZodString;
+}, z.core.$strip>;
 /**
- * Worker-relevant cluster settings a DB-less worker pulls from the leader.
+ * Worker-relevant cluster settings a DB-less worker pulls from a coordinator.
  *
  * A typed, concrete snapshot (not an open key/value bus): each worker-consumed
  * cluster knob is a field here. Adding the next worker-relevant knob is one
@@ -360,11 +542,12 @@ export declare const peerConfigReloadResponseSchema: z.ZodObject<{
  */
 export declare const workerClusterSettingsSchema: z.ZodObject<{
     agentTokenTtlMs: z.ZodNumber;
+    firecrackerApiSocketWaitMs: z.ZodOptional<z.ZodNumber>;
 }, z.core.$strip>;
 /**
- * Cluster-settings pull request: a DB-less worker asks the leader for the
- * current worker-settings snapshot after observing the leader's advertised
- * clusterSettingsVersion ahead of its own. The leader replies with
+ * Cluster-settings pull request: a DB-less worker asks a connected coordinator
+ * for the current worker-settings snapshot after a coordinator advertised a
+ * clusterSettingsVersion ahead of its own. The coordinator replies with
  * peer.clusterSettings.response.
  */
 export declare const peerClusterSettingsRequestSchema: z.ZodObject<{
@@ -372,7 +555,7 @@ export declare const peerClusterSettingsRequestSchema: z.ZodObject<{
     messageId: z.ZodString;
 }, z.core.$strip>;
 /**
- * Cluster-settings pull response: the leader resolves the snapshot from
+ * Cluster-settings pull response: the coordinator resolves the snapshot from
  * cluster_settings and replies with it plus the current version.
  */
 export declare const peerClusterSettingsResponseSchema: z.ZodObject<{
@@ -381,6 +564,7 @@ export declare const peerClusterSettingsResponseSchema: z.ZodObject<{
     version: z.ZodNumber;
     settings: z.ZodObject<{
         agentTokenTtlMs: z.ZodNumber;
+        firecrackerApiSocketWaitMs: z.ZodOptional<z.ZodNumber>;
     }, z.core.$strip>;
 }, z.core.$strip>;
 /** Which of THIS peer's downstream nodes to gather. all=true ignores the id lists. */
@@ -455,6 +639,7 @@ export declare const peerScalerEventSchema: z.ZodObject<{
     }>;
     detail: z.ZodString;
     timestampMs: z.ZodNumber;
+    final: z.ZodOptional<z.ZodBoolean>;
 }, z.core.$strip>;
 /** All peer-to-peer messages (outbound from this node). */
 export declare const peerToPeerMessageSchema: z.ZodDiscriminatedUnion<[z.ZodObject<{
@@ -588,6 +773,10 @@ export declare const peerToPeerMessageSchema: z.ZodDiscriminatedUnion<[z.ZodObje
     }, z.core.$strip>], "kind">>>;
     triedConnections: z.ZodArray<z.ZodString>;
     maxHops: z.ZodNumber;
+    spawnRetry: z.ZodOptional<z.ZodObject<{
+        maxAttempts: z.ZodNumber;
+        backoffMs: z.ZodNumber;
+    }, z.core.$strip>>;
     coordinatorId: z.ZodString;
     requestId: z.ZodOptional<z.ZodString>;
     traceId: z.ZodOptional<z.ZodString>;
@@ -718,6 +907,70 @@ export declare const peerToPeerMessageSchema: z.ZodDiscriminatedUnion<[z.ZodObje
     errors: z.ZodOptional<z.ZodArray<z.ZodString>>;
     restartRequired: z.ZodOptional<z.ZodArray<z.ZodString>>;
     fieldsChanged: z.ZodOptional<z.ZodArray<z.ZodString>>;
+}, z.core.$strip>, z.ZodObject<{
+    type: z.ZodLiteral<"peer.scaler.orphans.request">;
+    messageId: z.ZodString;
+    action: z.ZodEnum<{
+        list: "list";
+        stop: "stop";
+    }>;
+    vmIds: z.ZodOptional<z.ZodArray<z.ZodString>>;
+}, z.core.$strip>, z.ZodObject<{
+    type: z.ZodLiteral<"peer.scaler.orphans.response">;
+    messageId: z.ZodString;
+    ok: z.ZodBoolean;
+    error: z.ZodOptional<z.ZodString>;
+    firecrackerScalers: z.ZodOptional<z.ZodArray<z.ZodString>>;
+    vms: z.ZodOptional<z.ZodArray<z.ZodObject<{
+        vmId: z.ZodString;
+        scaler: z.ZodString;
+        pid: z.ZodNumber;
+        startedAt: z.ZodString;
+        ageSeconds: z.ZodNumber;
+        chrootDir: z.ZodString;
+        status: z.ZodEnum<{
+            orphaned: "orphaned";
+            tracked: "tracked";
+            unverified: "unverified";
+        }>;
+        trackedBy: z.ZodArray<z.ZodEnum<{
+            backend: "backend";
+            "bound-job": "bound-job";
+            registered: "registered";
+            spawning: "spawning";
+        }>>;
+        reason: z.ZodString;
+    }, z.core.$strip>>>;
+    results: z.ZodOptional<z.ZodArray<z.ZodObject<{
+        vmId: z.ZodString;
+        outcome: z.ZodEnum<{
+            error: "error";
+            "not-found": "not-found";
+            "not-live": "not-live";
+            stopped: "stopped";
+            tracked: "tracked";
+            unverified: "unverified";
+        }>;
+        pid: z.ZodOptional<z.ZodNumber>;
+        detail: z.ZodString;
+    }, z.core.$strip>>>;
+}, z.core.$strip>, z.ZodObject<{
+    type: z.ZodLiteral<"peer.forget.request">;
+    messageId: z.ZodString;
+    instanceId: z.ZodString;
+    acknowledgeBackstop: z.ZodOptional<z.ZodBoolean>;
+}, z.core.$strip>, z.ZodObject<{
+    type: z.ZodLiteral<"peer.forget.response">;
+    messageId: z.ZodString;
+    outcome: z.ZodEnum<{
+        "acknowledgement-required": "acknowledgement-required";
+        connected: "connected";
+        error: "error";
+        forgotten: "forgotten";
+        "not-found": "not-found";
+        recent: "recent";
+    }>;
+    detail: z.ZodString;
 }, z.core.$strip>, z.ZodObject<{
     type: z.ZodLiteral<"peer.clusterSettings.request">;
     messageId: z.ZodString;
@@ -727,6 +980,7 @@ export declare const peerToPeerMessageSchema: z.ZodDiscriminatedUnion<[z.ZodObje
     version: z.ZodNumber;
     settings: z.ZodObject<{
         agentTokenTtlMs: z.ZodNumber;
+        firecrackerApiSocketWaitMs: z.ZodOptional<z.ZodNumber>;
     }, z.core.$strip>;
 }, z.core.$strip>, z.ZodObject<{
     type: z.ZodLiteral<"peer.logs.collect.request">;
@@ -770,6 +1024,7 @@ export declare const peerToPeerMessageSchema: z.ZodDiscriminatedUnion<[z.ZodObje
     }>;
     detail: z.ZodString;
     timestampMs: z.ZodNumber;
+    final: z.ZodOptional<z.ZodBoolean>;
 }, z.core.$strip>], "type">;
 /** All peer-to-peer messages (inbound to this node). */
 export declare const peerFromPeerMessageSchema: z.ZodDiscriminatedUnion<[z.ZodObject<{
@@ -903,6 +1158,10 @@ export declare const peerFromPeerMessageSchema: z.ZodDiscriminatedUnion<[z.ZodOb
     }, z.core.$strip>], "kind">>>;
     triedConnections: z.ZodArray<z.ZodString>;
     maxHops: z.ZodNumber;
+    spawnRetry: z.ZodOptional<z.ZodObject<{
+        maxAttempts: z.ZodNumber;
+        backoffMs: z.ZodNumber;
+    }, z.core.$strip>>;
     coordinatorId: z.ZodString;
     requestId: z.ZodOptional<z.ZodString>;
     traceId: z.ZodOptional<z.ZodString>;
@@ -1033,6 +1292,70 @@ export declare const peerFromPeerMessageSchema: z.ZodDiscriminatedUnion<[z.ZodOb
     errors: z.ZodOptional<z.ZodArray<z.ZodString>>;
     restartRequired: z.ZodOptional<z.ZodArray<z.ZodString>>;
     fieldsChanged: z.ZodOptional<z.ZodArray<z.ZodString>>;
+}, z.core.$strip>, z.ZodObject<{
+    type: z.ZodLiteral<"peer.scaler.orphans.request">;
+    messageId: z.ZodString;
+    action: z.ZodEnum<{
+        list: "list";
+        stop: "stop";
+    }>;
+    vmIds: z.ZodOptional<z.ZodArray<z.ZodString>>;
+}, z.core.$strip>, z.ZodObject<{
+    type: z.ZodLiteral<"peer.scaler.orphans.response">;
+    messageId: z.ZodString;
+    ok: z.ZodBoolean;
+    error: z.ZodOptional<z.ZodString>;
+    firecrackerScalers: z.ZodOptional<z.ZodArray<z.ZodString>>;
+    vms: z.ZodOptional<z.ZodArray<z.ZodObject<{
+        vmId: z.ZodString;
+        scaler: z.ZodString;
+        pid: z.ZodNumber;
+        startedAt: z.ZodString;
+        ageSeconds: z.ZodNumber;
+        chrootDir: z.ZodString;
+        status: z.ZodEnum<{
+            orphaned: "orphaned";
+            tracked: "tracked";
+            unverified: "unverified";
+        }>;
+        trackedBy: z.ZodArray<z.ZodEnum<{
+            backend: "backend";
+            "bound-job": "bound-job";
+            registered: "registered";
+            spawning: "spawning";
+        }>>;
+        reason: z.ZodString;
+    }, z.core.$strip>>>;
+    results: z.ZodOptional<z.ZodArray<z.ZodObject<{
+        vmId: z.ZodString;
+        outcome: z.ZodEnum<{
+            error: "error";
+            "not-found": "not-found";
+            "not-live": "not-live";
+            stopped: "stopped";
+            tracked: "tracked";
+            unverified: "unverified";
+        }>;
+        pid: z.ZodOptional<z.ZodNumber>;
+        detail: z.ZodString;
+    }, z.core.$strip>>>;
+}, z.core.$strip>, z.ZodObject<{
+    type: z.ZodLiteral<"peer.forget.request">;
+    messageId: z.ZodString;
+    instanceId: z.ZodString;
+    acknowledgeBackstop: z.ZodOptional<z.ZodBoolean>;
+}, z.core.$strip>, z.ZodObject<{
+    type: z.ZodLiteral<"peer.forget.response">;
+    messageId: z.ZodString;
+    outcome: z.ZodEnum<{
+        "acknowledgement-required": "acknowledgement-required";
+        connected: "connected";
+        error: "error";
+        forgotten: "forgotten";
+        "not-found": "not-found";
+        recent: "recent";
+    }>;
+    detail: z.ZodString;
 }, z.core.$strip>, z.ZodObject<{
     type: z.ZodLiteral<"peer.clusterSettings.request">;
     messageId: z.ZodString;
@@ -1042,6 +1365,7 @@ export declare const peerFromPeerMessageSchema: z.ZodDiscriminatedUnion<[z.ZodOb
     version: z.ZodNumber;
     settings: z.ZodObject<{
         agentTokenTtlMs: z.ZodNumber;
+        firecrackerApiSocketWaitMs: z.ZodOptional<z.ZodNumber>;
     }, z.core.$strip>;
 }, z.core.$strip>, z.ZodObject<{
     type: z.ZodLiteral<"peer.logs.collect.request">;
@@ -1085,11 +1409,13 @@ export declare const peerFromPeerMessageSchema: z.ZodDiscriminatedUnion<[z.ZodOb
     }>;
     detail: z.ZodString;
     timestampMs: z.ZodNumber;
+    final: z.ZodOptional<z.ZodBoolean>;
 }, z.core.$strip>], "type">;
 export type PeerCapabilities = z.infer<typeof peerCapabilitiesSchema>;
 export type ScalerCapacitySummary = z.infer<typeof scalerCapacitySummarySchema>;
 export type PeerHeartbeat = z.infer<typeof peerHeartbeatSchema>;
 export type JobReroute = z.infer<typeof jobRerouteSchema>;
+export type RerouteSpawnRetry = z.infer<typeof rerouteSpawnRetrySchema>;
 export type JobProgress = z.infer<typeof jobProgressSchema>;
 export type JobProgressAck = z.infer<typeof jobProgressAckSchema>;
 export type PeerScalerEvent = z.infer<typeof peerScalerEventSchema>;
@@ -1102,6 +1428,10 @@ export type PeerCacheUploadRequest = z.infer<typeof peerCacheUploadRequestSchema
 export type PeerCacheUploadResponse = z.infer<typeof peerCacheUploadResponseSchema>;
 export type PeerConfigReload = z.infer<typeof peerConfigReloadSchema>;
 export type PeerConfigReloadResponse = z.infer<typeof peerConfigReloadResponseSchema>;
+export type PeerScalerOrphansRequest = z.infer<typeof peerScalerOrphansRequestSchema>;
+export type PeerScalerOrphansResponse = z.infer<typeof peerScalerOrphansResponseSchema>;
+export type PeerForgetRequest = z.infer<typeof peerForgetRequestSchema>;
+export type PeerForgetResponse = z.infer<typeof peerForgetResponseSchema>;
 export type WorkerClusterSettings = z.infer<typeof workerClusterSettingsSchema>;
 export type PeerClusterSettingsRequest = z.infer<typeof peerClusterSettingsRequestSchema>;
 export type PeerClusterSettingsResponse = z.infer<typeof peerClusterSettingsResponseSchema>;

package/dist/protocol/messages/peer.js

@@ -126,7 +126,7 @@ const peerHeartbeatSchema = z.object({
 	configVersion: z.number().optional(),
 	/** Registry version for cross-orchestrator registration sync (backward compatible). */
 	registryVersion: z.number().optional(),
-	/** Shared cluster-settings version for leader→worker settings pull (backward compatible). */
+	/** Cluster-settings version for the coordinator→worker settings pull (backward compatible). */
 	clusterSettingsVersion: z.number().optional(),
 	timestamp: z.number(),
 	hostname: z.string().optional(),
@@ -141,6 +141,15 @@ const peerHeartbeatSchema = z.object({
 	runningAsUid: z.number().nullable().optional(),
 	version: z.string().optional()
 });
+/**
+* The spawn-retry budget a worker applies to one rerouted job: how many agent
+* spawns it attempts, and how long it waits after a failed one. The sending
+* coordinator resolves it per org and enforces the same budget on its side.
+*/
+const rerouteSpawnRetrySchema = z.object({
+	maxAttempts: z.number().int().min(1),
+	backoffMs: z.number().int().min(0)
+});
 /** Request to reroute a job to another orchestrator (no local agent can handle it). */
 const jobRerouteSchema = z.object({
 	type: z.literal("job.reroute"),
@@ -178,6 +187,11 @@ const jobRerouteSchema = z.object({
 	excludePatterns: z.array(LabelMatcher).optional(),
 	triedConnections: z.array(z.string()),
 	maxHops: z.number(),
+	/**
+	* The worker's spawn-retry budget for this job. Absent from an older sender:
+	* the worker applies its own configured default.
+	*/
+	spawnRetry: rerouteSpawnRetrySchema.optional(),
 	coordinatorId: z.string(),
 	requestId: z.string().optional(),
 	traceId: z.string().optional(),
@@ -346,18 +360,135 @@ const peerConfigReloadResponseSchema = z.object({
 	restartRequired: z.array(z.string()).optional(),
 	fieldsChanged: z.array(z.string()).optional()
 });
+/** How a live Firecracker VM relates to the orchestrator on its host. */
+const ScalerVmStatus = z.enum([
+	"orphaned",
+	"unverified",
+	"tracked"
+]);
+/** What tracks a live VM. `bound-job` is answered from a coordinator's dispatch queue. */
+const ScalerVmTracker = z.enum([
+	"backend",
+	"spawning",
+	"registered",
+	"bound-job"
+]);
+/** What a stop did to one VM. Only `stopped` signalled anything. */
+const ScalerVmStopOutcome = z.enum([
+	"stopped",
+	"tracked",
+	"unverified",
+	"not-live",
+	"not-found",
+	"error"
+]);
+/** What a scaler orphan request asks the target node to do. */
+const ScalerOrphansAction = z.enum(["list", "stop"]);
+/** One live Firecracker VM on a node, as its orchestrator sees it. */
+const scalerLiveVmSchema = z.object({
+	vmId: z.string(),
+	scaler: z.string(),
+	pid: z.number().int().positive(),
+	startedAt: z.string(),
+	ageSeconds: z.number().nonnegative(),
+	chrootDir: z.string(),
+	status: ScalerVmStatus,
+	trackedBy: z.array(ScalerVmTracker),
+	reason: z.string()
+});
+/** The result of stopping one VM. `pid` is the process that was (or would be) signalled. */
+const scalerVmStopResultSchema = z.object({
+	vmId: z.string(),
+	outcome: ScalerVmStopOutcome,
+	pid: z.number().int().positive().optional(),
+	detail: z.string()
+});
+/**
+* Scaler orphan request: forwarded by a coordinator to the node an operator
+* targeted with `kici-admin scaler orphans --target`. The node answers from
+* its own host and its own tracking with peer.scaler.orphans.response.
+*/
+const peerScalerOrphansRequestSchema = z.object({
+	type: z.literal("peer.scaler.orphans.request"),
+	messageId: z.string(),
+	action: ScalerOrphansAction,
+	/** `stop` only: the VM ids the operator approved. */
+	vmIds: z.array(z.string()).optional()
+});
+/**
+* Scaler orphan response. `vms` answers a `list`, `results` answers a `stop`;
+* `ok: false` carries the node's `error` instead.
+*/
+const peerScalerOrphansResponseSchema = z.object({
+	type: z.literal("peer.scaler.orphans.response"),
+	messageId: z.string(),
+	ok: z.boolean(),
+	error: z.string().optional(),
+	/** The node's Firecracker scaler names; empty when it runs none. */
+	firecrackerScalers: z.array(z.string()).optional(),
+	vms: z.array(scalerLiveVmSchema).optional(),
+	results: z.array(scalerVmStopResultSchema).optional()
+});
 /**
-* Worker-relevant cluster settings a DB-less worker pulls from the leader.
+* What forgetting a departed peer did on one coordinator. `recent`: the peer
+* was heard from inside the window this coordinator still treats it as alive
+* for (the longer of the stale window and the reroute flap grace, or the
+* backstop grace of a peer that adopted provisions), so it may be partitioned
+* rather than gone. `acknowledgement-required`: forgetting it would switch this
+* coordinator's event-provision backstop back on, and the request did not
+* acknowledge that. Every outcome but `forgotten` keeps the peer.
+*/
+const PeerForgetOutcome = z.enum([
+	"forgotten",
+	"not-found",
+	"connected",
+	"recent",
+	"acknowledgement-required",
+	"error"
+]);
+/**
+* Forget request: a coordinator that forgot a departed peer (`kici-admin peer
+* forget`) asks each connected sibling coordinator to drop it from its own
+* live peer registry too.
+*/
+const peerForgetRequestSchema = z.object({
+	type: z.literal("peer.forget.request"),
+	messageId: z.string(),
+	/** The departed peer's instance id. */
+	instanceId: z.string(),
+	/**
+	* The operator acknowledged that forgetting the peer may switch the
+	* event-provision backstop back on. Absent means not acknowledged.
+	*/
+	acknowledgeBackstop: z.boolean().optional()
+});
+/** Forget response: what the sibling did with the request. */
+const peerForgetResponseSchema = z.object({
+	type: z.literal("peer.forget.response"),
+	messageId: z.string(),
+	outcome: PeerForgetOutcome,
+	detail: z.string()
+});
+/**
+* Worker-relevant cluster settings a DB-less worker pulls from a coordinator.
 *
 * A typed, concrete snapshot (not an open key/value bus): each worker-consumed
 * cluster knob is a field here. Adding the next worker-relevant knob is one
 * field on this schema, not a new protocol message.
 */
-const workerClusterSettingsSchema = z.object({ agentTokenTtlMs: z.number() });
+const workerClusterSettingsSchema = z.object({
+	agentTokenTtlMs: z.number(),
+	/**
+	* How long a Firecracker spawn waits for the VM's API socket. Optional: a
+	* leader that predates the knob omits it, and the worker keeps its own
+	* configured default.
+	*/
+	firecrackerApiSocketWaitMs: z.number().optional()
+});
 /**
-* Cluster-settings pull request: a DB-less worker asks the leader for the
-* current worker-settings snapshot after observing the leader's advertised
-* clusterSettingsVersion ahead of its own. The leader replies with
+* Cluster-settings pull request: a DB-less worker asks a connected coordinator
+* for the current worker-settings snapshot after a coordinator advertised a
+* clusterSettingsVersion ahead of its own. The coordinator replies with
 * peer.clusterSettings.response.
 */
 const peerClusterSettingsRequestSchema = z.object({
@@ -365,7 +496,7 @@ const peerClusterSettingsRequestSchema = z.object({
 	messageId: z.string()
 });
 /**
-* Cluster-settings pull response: the leader resolves the snapshot from
+* Cluster-settings pull response: the coordinator resolves the snapshot from
 * cluster_settings and replies with it plus the current version.
 */
 const peerClusterSettingsResponseSchema = z.object({
@@ -443,7 +574,14 @@ const peerScalerEventSchema = z.object({
 	/** Human-readable detail, including any captured spawn stderr tail. */
 	detail: z.string(),
 	/** Event timestamp in epoch milliseconds. */
-	timestampMs: z.number()
+	timestampMs: z.number(),
+	/**
+	* The worker's retry verdict on a `scaler.failed` for a job under its spawn-retry
+	* budget: false while attempts remain, true on the last one. Absent on a repeated
+	* report of one failed spawn, on every other relay, and from an older worker; the
+	* coordinator then keeps its spawn window running unchanged.
+	*/
+	final: z.boolean().optional()
 });
 /** All peer-to-peer messages (outbound from this node). */
 const peerToPeerMessageSchema = z.discriminatedUnion("type", [
@@ -465,6 +603,10 @@ const peerToPeerMessageSchema = z.discriminatedUnion("type", [
 	peerCacheUploadResponseSchema,
 	peerConfigReloadSchema,
 	peerConfigReloadResponseSchema,
+	peerScalerOrphansRequestSchema,
+	peerScalerOrphansResponseSchema,
+	peerForgetRequestSchema,
+	peerForgetResponseSchema,
 	peerClusterSettingsRequestSchema,
 	peerClusterSettingsResponseSchema,
 	peerLogsCollectRequestSchema,
@@ -494,6 +636,10 @@ const peerFromPeerMessageSchema = z.discriminatedUnion("type", [
 	peerCacheUploadResponseSchema,
 	peerConfigReloadSchema,
 	peerConfigReloadResponseSchema,
+	peerScalerOrphansRequestSchema,
+	peerScalerOrphansResponseSchema,
+	peerForgetRequestSchema,
+	peerForgetResponseSchema,
 	peerClusterSettingsRequestSchema,
 	peerClusterSettingsResponseSchema,
 	peerLogsCollectRequestSchema,
@@ -504,6 +650,6 @@ const peerFromPeerMessageSchema = z.discriminatedUnion("type", [
 	peerScalerEventSchema
 ]);
 //#endregion
-export { fleetSelectionSchema, jobProgressAckSchema, jobProgressSchema, jobRerouteAckSchema, jobRerouteSchema, peerAgentTokenRevokeSchema, peerAuthRequestSchema, peerAuthResponseSchema, peerCapabilitiesSchema, peerClusterSettingsRequestSchema, peerClusterSettingsResponseSchema, peerConfigReloadResponseSchema, peerConfigReloadSchema, peerFromPeerMessageSchema, peerHeartbeatSchema, peerHelloResponseSchema, peerHelloSchema, peerJobCancelSchema, peerLeavingSchema, peerLogsCollectChunkSchema, peerLogsCollectErrorSchema, peerLogsCollectRequestSchema, peerScalerEventSchema, peerToPeerMessageSchema, raftAppendEntriesSchema, raftVoteRequestSchema, raftVoteResponseSchema, workerClusterSettingsSchema };
+export { PeerForgetOutcome, ScalerOrphansAction, ScalerVmStatus, ScalerVmStopOutcome, ScalerVmTracker, fleetSelectionSchema, jobProgressAckSchema, jobProgressSchema, jobRerouteAckSchema, jobRerouteSchema, peerAgentTokenRevokeSchema, peerAuthRequestSchema, peerAuthResponseSchema, peerCapabilitiesSchema, peerClusterSettingsRequestSchema, peerClusterSettingsResponseSchema, peerConfigReloadResponseSchema, peerConfigReloadSchema, peerForgetRequestSchema, peerForgetResponseSchema, peerFromPeerMessageSchema, peerHeartbeatSchema, peerHelloResponseSchema, peerHelloSchema, peerJobCancelSchema, peerLeavingSchema, peerLogsCollectChunkSchema, peerLogsCollectErrorSchema, peerLogsCollectRequestSchema, peerScalerEventSchema, peerScalerOrphansRequestSchema, peerScalerOrphansResponseSchema, peerToPeerMessageSchema, raftAppendEntriesSchema, raftVoteRequestSchema, raftVoteResponseSchema, rerouteSpawnRetrySchema, scalerLiveVmSchema, scalerVmStopResultSchema, workerClusterSettingsSchema };
 
 //# sourceMappingURL=peer.js.map