@voltro/workflow 0.84.0 → 0.86.0

package/dist/cluster.js

@@ -1,2 +1,8 @@
-import { _ as e, a as t, c as n, d as r, f as i, g as a, h as o, i as s, l as c, m as l, n as u, o as d, p as f, r as p, s as m, t as h, u as g, v as _ } from "./cluster-i3frC3Da.js";
-export { u as DEFAULT_RUNNER_LISTEN_PORT, f as WorkflowRestartDispatcher, _ as WorkflowResumeDispatcher, l as activityPrimaryKey, i as clusterMigrationGateLayer, p as detectWsrepCluster, h as makeClusterCronLayer, r as makeClusterDiagnostics, s as parseShardLockMode, t as pollWorkflowStorage, o as readWorkflowRunReply, a as redriveFailedRun, d as resolveFailoverTuning, m as resolveRunnerIdentity, n as resolveRunnerStorage, c as resolveShardLockDisableAdvisory, e as resumeRunFromStep, g as workflowEngineLayer };
+import { WorkflowResumeDispatcher as e } from "./workflowResumeDispatch.js";
+import { JournalExit as t, activityPrimaryKey as n, readWorkflowRunReply as r, redriveFailedRun as i, resumeRunFromStep as a } from "./workflowJournal.js";
+import { WorkflowRestartDispatcher as o } from "./workflowRestartDispatch.js";
+import { clusterMigrationGateLayer as s } from "./clusterMigrationGate.js";
+import { clusterLockNamespaceOid as c, makeClusterDiagnostics as l } from "./clusterDiagnosticsSql.js";
+import { DEFAULT_RUNNER_LISTEN_PORT as u, detectWsrepCluster as d, parseShardLockMode as f, pollWorkflowStorage as p, resolveFailoverTuning as m, resolveRunnerIdentity as h, resolveRunnerStorage as g, resolveShardLockDisableAdvisory as _, workflowEngineLayer as v } from "./clusterLayer.js";
+import { makeClusterCronLayer as y } from "./clusterCron.js";
+export { u as DEFAULT_RUNNER_LISTEN_PORT, t as JournalExit, o as WorkflowRestartDispatcher, e as WorkflowResumeDispatcher, n as activityPrimaryKey, c as clusterLockNamespaceOid, s as clusterMigrationGateLayer, d as detectWsrepCluster, y as makeClusterCronLayer, l as makeClusterDiagnostics, f as parseShardLockMode, p as pollWorkflowStorage, r as readWorkflowRunReply, i as redriveFailedRun, m as resolveFailoverTuning, h as resolveRunnerIdentity, g as resolveRunnerStorage, _ as resolveShardLockDisableAdvisory, a as resumeRunFromStep, v as workflowEngineLayer };

package/dist/clusterCron.js

@@ -0,0 +1,11 @@
+import { Duration as e, Effect as t } from "effect";
+import { ClusterCron as n } from "effect/cluster";
+//#region src/clusterCron.ts
+var r = (r) => n.make({
+	name: r.name,
+	cron: r.cron,
+	execute: t.promise(() => r.execute()),
+	skipIfOlderThan: r.skipIfOlderThan ?? e.days(1)
+});
+//#endregion
+export { r as makeClusterCronLayer };

package/dist/clusterDiagnostics.js

@@ -0,0 +1,20 @@
+import { Context as e, Effect as t } from "effect";
+//#region src/clusterDiagnostics.ts
+var n = 6e5, r = class extends e.Service()("@voltro/workflow/ClusterDiagnostics") {}, i = (e, n, r = Date.now()) => t.flatMap(e.messagesFor(n), (n) => {
+	if (n.length === 0) return t.succeed(void 0);
+	let i = [...new Set(n.filter((e) => !e.processed).map((e) => e.shardId))], a = i.length === 0 ? t.succeed([]) : e.shardLockHolders(i);
+	return t.map(a, (t) => ({
+		runner: e.ownAddress,
+		shardLockMode: e.shardLockMode,
+		messages: n.map((e) => ({
+			tag: e.tag,
+			shardId: e.shardId,
+			processed: e.processed,
+			lastRead: e.lastRead === null ? null : e.lastRead.toISOString(),
+			unreadForMs: !e.processed && e.lastRead !== null ? Math.max(0, r - e.lastRead.getTime()) : null
+		})),
+		shardLocks: t
+	}));
+});
+//#endregion
+export { n as CLUSTER_REREAD_WINDOW_MS, r as ClusterDiagnostics, i as describeStalledRun };

package/dist/clusterDiagnosticsSql.js

@@ -0,0 +1,145 @@
+import { CLUSTER_REREAD_WINDOW_MS as e } from "./clusterDiagnostics.js";
+import { Effect as t } from "effect";
+import { SqlClient as n } from "effect/sql";
+//#region src/clusterDiagnosticsSql.ts
+var r = (e) => {
+	if (e instanceof Date) return Number.isNaN(e.getTime()) ? null : e;
+	if (typeof e == "string" || typeof e == "number") {
+		let t = typeof e == "string" && /^\d{4}-\d{2}-\d{2} \d{2}:\d{2}:\d{2}$/.test(e) ? `${e.replace(" ", "T")}Z` : e, n = new Date(t);
+		return Number.isNaN(n.getTime()) ? null : n;
+	}
+	return null;
+}, i = (e) => e === !0 || e === 1 || e === "1" || e instanceof Uint8Array && e[0] === 1, a = "cluster", o = (e = a) => {
+	let t = new TextEncoder().encode(`effect-cluster:${e}`), n = 2166136261;
+	for (let e of t) n = Math.imul(n ^ e, 16777619);
+	return (n | 0) >>> 0;
+}, s = (e, t) => {
+	let n = e.lastIndexOf(":");
+	if (n < 0) return;
+	let r = e.slice(0, n), i = Number(e.slice(n + 1)), a = t.indexOf(r);
+	if (!(a < 0 || !Number.isInteger(i))) return (a + 1) * 1e6 + i;
+}, c = (a) => t.gen(function* () {
+	let c = yield* n.SqlClient, l = a.shardGroups ?? ["default"], u = a.dialectId ?? "", d = u.startsWith("postgres") || u === "pg", f = (e) => e.pipe(t.catch(() => t.succeed([]))), p = c.onDialectOrElse({
+		pg: () => c.literal("NOW()"),
+		mysql: () => c.literal("NOW()"),
+		mssql: () => c.literal("GETDATE()"),
+		orElse: () => c.literal("CURRENT_TIMESTAMP")
+	}), m = (e, t) => {
+		let n = r(e.last_read), a = r(e.server_now), o = n !== null && a !== null ? Math.max(0, a.getTime() - n.getTime()) : null;
+		return {
+			tag: typeof e.tag == "string" ? e.tag : null,
+			entityType: String(e.entity_type ?? ""),
+			entityId: String(e.entity_id ?? ""),
+			shardId: String(e.shard_id ?? ""),
+			processed: i(e.processed),
+			lastRead: o === null ? null : new Date(t - o)
+		};
+	}, h = (e) => c.onDialectOrElse({
+		pg: () => c.literal(`NOW() - INTERVAL '${e} seconds'`),
+		mysql: () => c.literal(`NOW() - INTERVAL ${e} SECOND`),
+		mssql: () => c.literal(`DATEADD(SECOND, -${e}, GETDATE())`),
+		orElse: () => c.literal(`DATETIME('now', '-${e} seconds')`)
+	}), g = c.onDialectOrElse({
+		pg: () => c.literal("FALSE"),
+		orElse: () => c.literal("0")
+	}), _ = (e) => {
+		let n = Date.now();
+		return f(c`
+        SELECT tag, entity_type, entity_id, shard_id, processed, last_read, ${p} AS server_now
+        FROM cluster_messages
+        WHERE entity_id = ${e}
+        ORDER BY rowid ASC
+      `).pipe(t.map((e) => e.map((e) => m(e, n))));
+	}, v = (n = e) => {
+		let r = Math.max(1, Math.ceil(n / 1e3)), i = Date.now();
+		return f(c`
+        SELECT tag, entity_type, entity_id, shard_id, processed, last_read, ${p} AS server_now
+        FROM cluster_messages
+        WHERE processed = ${g}
+        AND last_read IS NOT NULL
+        AND last_read < ${h(r)}
+        AND (deliver_at IS NULL OR deliver_at <= ${i})
+        ORDER BY last_read ASC
+      `).pipe(t.map((e) => e.flatMap((e) => {
+			let t = m(e, i);
+			return t.lastRead === null ? [] : [{
+				...t,
+				unreadForMs: Math.max(0, i - t.lastRead.getTime())
+			}];
+		})));
+	}, y = (e) => f(c`
+        SELECT shard_id, address, acquired_at FROM cluster_locks WHERE ${c.in("shard_id", e)}
+      `).pipe(t.map((e) => e.map((e) => {
+		let t = e, n = String(t.address ?? ""), i = String(t.shard_id ?? "");
+		return {
+			shardId: i,
+			mode: "row",
+			holder: n,
+			own: n === a.ownAddress || (a.hasShard?.(i) ?? !1),
+			since: r(t.acquired_at)
+		};
+	}))), b = (e) => {
+		let n = /* @__PURE__ */ new Map();
+		for (let t of e) {
+			let e = s(t, l);
+			e !== void 0 && n.set(e, t);
+		}
+		return n.size === 0 ? t.succeed([]) : t.gen(function* () {
+			return (yield* f(c`
+          SELECT l.objid, a.pid, a.client_addr, a.application_name, a.state, a.state_change, a.backend_start, a.wait_event_type
+          FROM pg_locks l
+          JOIN pg_stat_activity a ON a.pid = l.pid
+          WHERE l.locktype = 'advisory' AND l.granted AND l.objsubid = 2
+            AND l.classid = ${o()} AND ${c.in("l.objid", [...n.keys()])}
+        `)).flatMap((e) => {
+				let t = e, i = n.get(Number(t.objid));
+				if (i === void 0) return [];
+				let o = Number(t.pid), s = t.client_addr === null || t.client_addr === void 0 ? "local socket" : String(t.client_addr), c = a.hasShard?.(i) ?? !1;
+				return [{
+					shardId: i,
+					mode: "advisory",
+					holder: c ? `this runner (pid ${o})` : `pid ${o} from ${s}`,
+					own: c,
+					since: r(t.state_change),
+					detail: {
+						pid: o,
+						clientAddr: t.client_addr ?? null,
+						applicationName: t.application_name ?? null,
+						state: t.state ?? null,
+						waitEventType: t.wait_event_type ?? null,
+						backendStart: r(t.backend_start)?.toISOString() ?? null
+					}
+				}];
+			});
+		});
+	};
+	return {
+		ownAddress: a.ownAddress,
+		shardLockMode: a.shardLockMode,
+		messagesFor: _,
+		unreadMessages: v,
+		shardLockHolders: (e) => e.length === 0 || a.runnerStorage !== "sql" ? t.succeed([]) : a.shardLockMode === "row" ? y(e) : a.shardLockMode === "advisory" && d ? b(e) : t.succeed([]),
+		runners: () => a.runnerStorage === "sql" ? f(c`SELECT address, last_heartbeat, healthy FROM cluster_runners ORDER BY last_heartbeat DESC`).pipe(t.map((e) => e.map((e) => {
+			let t = e;
+			return {
+				address: String(t.address ?? ""),
+				lastHeartbeat: r(t.last_heartbeat),
+				healthy: i(t.healthy)
+			};
+		}))) : t.succeed([]),
+		discardMessages: (e) => {
+			let r = e.unreadForMs === void 0 ? c.literal("") : c`AND last_read IS NOT NULL AND last_read < ${h(Math.max(1, Math.ceil(e.unreadForMs / 1e3)))}`;
+			return t.gen(function* () {
+				let t = (yield* c`
+          SELECT id FROM cluster_messages
+          WHERE entity_type = ${e.entityType} AND entity_id = ${e.entityId}
+          AND processed = ${g}
+          ${r}
+        `).map((e) => e.id).filter((e) => e != null).map(String);
+				return t.length === 0 ? 0 : (yield* c`DELETE FROM cluster_replies WHERE ${c.in("request_id", t)}`, yield* c`DELETE FROM cluster_messages WHERE ${c.in("id", t)}`, t.length);
+			}).pipe(c.withTransaction, t.provideService(n.SafeIntegers, !0), t.catch(() => t.succeed(0)));
+		}
+	};
+});
+//#endregion
+export { o as clusterLockNamespaceOid, c as makeClusterDiagnostics };

package/dist/clusterLayer.js

@@ -0,0 +1,110 @@
+import { withWorkflowDelivery as e } from "./workflowDelivery.js";
+import { ClusterDiagnostics as t } from "./clusterDiagnostics.js";
+import { WorkflowResumeDispatcher as n, makeWorkflowResumeDispatcher as r } from "./workflowResumeDispatch.js";
+import { WorkflowRestartDispatcher as i, makeWorkflowRestartDispatcher as a } from "./workflowRestartDispatch.js";
+import { clusterMigrationGateLayer as o } from "./clusterMigrationGate.js";
+import { clusterRunnerLayer as s } from "./clusterRunnerLayer.js";
+import { acquireSqliteWorkflowOwnership as c } from "./sqliteWorkflowOwnership.js";
+import { makeClusterDiagnostics as l } from "./clusterDiagnosticsSql.js";
+import { RateLimiter as u } from "effect/persistence";
+import { Context as d, Duration as f, Effect as p, Layer as m, Option as h } from "effect";
+import { SqlClient as g } from "effect/sql";
+import { ClusterWorkflowEngine as _, MessageStorage as v, RunnerAddress as y, ShardId as b, Sharding as x } from "effect/cluster";
+import { createLogger as S } from "@voltro/logger";
+//#region src/clusterLayer.ts
+var C = S({ scope: "voltro:workflow" }), w = 34e3, T = (e) => {
+	let t = e.runnerListenHost ?? process.env.POD_IP ?? process.env.VOLTRO_WORKFLOW_RUNNER_HOST ?? "localhost";
+	return {
+		host: t,
+		port: e.runnerListenPort ?? 34e3,
+		localhostRisk: e.runnerStorage === "sql" && (t === "localhost" || t === "127.0.0.1")
+	};
+}, E = (e) => {
+	let t = e.sqlRunnerSupported ?? !e.sqliteFamily, n = t ? "sql" : "memory", r = e.override ?? process.env.VOLTRO_WORKFLOW_RUNNER_STORAGE;
+	if (r === void 0 || r.trim() === "") return n;
+	let i = r.trim().toLowerCase();
+	if (i !== "memory" && i !== "sql") throw Error(`VOLTRO_WORKFLOW_RUNNER_STORAGE must be 'memory' or 'sql', got '${r}'.`);
+	if (i === "sql" && !t) throw Error("VOLTRO_WORKFLOW_RUNNER_STORAGE='sql' is unsupported for this database connection. Use 'memory' for a local or embedded SQLite-family connection.");
+	return i;
+}, D = (e = process.env.VOLTRO_WORKFLOW_SHARD_LOCK) => {
+	if (e === void 0 || e.trim() === "") return "auto";
+	let t = e.trim().toLowerCase();
+	if (t === "auto" || t === "row" || t === "advisory") return t;
+	throw Error(`VOLTRO_WORKFLOW_SHARD_LOCK must be 'auto', 'row', or 'advisory', got '${e}'.`);
+}, O = p.gen(function* () {
+	let e = yield* (yield* g.SqlClient)`SELECT @@wsrep_on AS wsrep`.pipe(p.catch(() => p.succeed([])));
+	if (e.length === 0) return !1;
+	let t = e[0].wsrep;
+	return t === 1 || t === "1" || t === !0 || t === "ON" || t === "on";
+}), k = (e) => p.gen(function* () {
+	return e.runnerStorage === "memory" ? !1 : e.dialectId === "sqlite" || e.dialectId === "turso" || e.mode === "row" ? !0 : e.mode === "advisory" || e.dialectId !== "mysql" && e.dialectId !== "mariadb" ? !1 : yield* O;
+}), A = (e) => ({
+	coordination: e.disableAdvisory ? "row-based" : "advisory",
+	dialect: e.dialectId ?? "unknown",
+	mode: e.mode,
+	...e.disableAdvisory && e.mode === "auto" ? { reason: e.dialectId === "sqlite" || e.dialectId === "turso" ? "SQLite-family connection uses row locks" : "wsrep/Galera cluster detected" } : {}
+}), j = () => {
+	let e = (e) => {
+		if (e === void 0) return;
+		let t = Number(e);
+		return Number.isFinite(t) && t > 0 ? t : void 0;
+	}, t = e(process.env.VOLTRO_WORKFLOW_FAILOVER_LEASE), n = e(process.env.VOLTRO_WORKFLOW_FAILOVER_HEARTBEAT), r = e(process.env.VOLTRO_WORKFLOW_POLL_INTERVAL);
+	return {
+		...t === void 0 ? {} : { failoverLeaseSeconds: t },
+		...n === void 0 ? {} : { failoverHeartbeatSeconds: n },
+		...r === void 0 ? {} : { messagePollSeconds: r }
+	};
+}, M = p.flatMap(x.Sharding, (e) => e.pollStorage), N = /* @__PURE__ */ new Set(), P = (e) => Object.assign(Object.create(Object.getPrototypeOf(e)), e, { requestIdForPrimaryKey: (t) => e.requestIdForPrimaryKey(t).pipe(p.provideService(g.SafeIntegers, !0)) }), F = (S) => {
+	let w = T({
+		runnerStorage: S.runnerStorage,
+		runnerListenHost: S.runnerListenHost,
+		runnerListenPort: S.runnerListenPort
+	}), E = w.host, O = w.port;
+	w.localhostRisk && !N.has(E) && (N.add(E), C.warn("runner host is not routable from other pods with SQL cluster storage — a workflow cannot resume on another pod after a reschedule. Inject POD_IP via the K8s downward API (fieldRef: status.podIP) or set runnerListenHost / VOLTRO_WORKFLOW_RUNNER_HOST.", { runnerHost: E }));
+	let j = S.dialectId === "sqlite" ? m.effect(g.SqlClient, p.gen(function* () {
+		let e = yield* g.SqlClient, t = (yield* e.unsafe("PRAGMA database_list")).find((e) => e.name === "main");
+		return t === void 0 ? yield* p.die(/* @__PURE__ */ Error("SQLite workflow connection has no main database")) : (t.file !== "" && (yield* c(t.file)), e);
+	}).pipe(p.orDie)).pipe(m.provide(S.sqlClientLayer)) : S.sqlClientLayer, M = o({
+		runnerStorage: S.runnerStorage,
+		charset: S.charset
+	}).pipe(m.provide(j)), F = S.shardLockMode ?? D(), I = m.unwrap(k({
+		dialectId: S.dialectId,
+		runnerStorage: S.runnerStorage,
+		mode: F
+	}).pipe(p.map((e) => {
+		S.runnerStorage === "sql" && C.info("shard-lock coordination", A({
+			disableAdvisory: e,
+			dialectId: S.dialectId,
+			mode: F
+		}));
+		let o = m.effect(t, p.flatMap(x.Sharding, (t) => l({
+			ownAddress: y.make(E, O).toString(),
+			shardLockMode: S.runnerStorage === "sql" ? e ? "row" : "advisory" : "none",
+			runnerStorage: S.runnerStorage,
+			dialectId: S.dialectId,
+			hasShard: (e) => t.hasShardId(b.fromString(e))
+		}))), c = {
+			runnerAddress: h.some(y.make(E, O)),
+			runnerListenAddress: h.some(y.make(E, O)),
+			...e ? { shardLockDisableAdvisory: !0 } : {},
+			...S.failoverLeaseSeconds === void 0 ? {} : { shardLockExpiration: f.seconds(S.failoverLeaseSeconds) },
+			...S.failoverHeartbeatSeconds === void 0 ? {} : { shardLockRefreshInterval: f.seconds(S.failoverHeartbeatSeconds) },
+			...S.messagePollSeconds === void 0 ? {} : { entityMessagePollInterval: f.seconds(S.messagePollSeconds) }
+		}, u = f.seconds(S.failoverHeartbeatSeconds ?? 10), d = s({
+			runnerStorage: S.runnerStorage,
+			shardingConfig: c,
+			guard: {
+				interval: u,
+				giveUpAfter: f.times(u, 3),
+				log: C
+			}
+		});
+		return m.mergeAll(o, m.effect(n, r), m.effect(i, a)).pipe(m.provideMerge(d));
+	}))).pipe(m.provide(M)), L = S.extraShardingLayers ?? [], R = L.length === 0 ? _.layer : m.mergeAll(_.layer, ...L), z = m.flatMap(I, (t) => {
+		let n = d.add(t, x.Sharding, e(d.get(t, x.Sharding)));
+		return m.succeedContext(S.dialectId === "sqlite" ? d.add(n, v.MessageStorage, P(d.get(n, v.MessageStorage))) : n);
+	}), B = u.layer.pipe(m.provide(u.layerStoreMemory));
+	return m.merge(R, B).pipe(m.provideMerge(z));
+};
+//#endregion
+export { w as DEFAULT_RUNNER_LISTEN_PORT, O as detectWsrepCluster, D as parseShardLockMode, M as pollWorkflowStorage, j as resolveFailoverTuning, T as resolveRunnerIdentity, E as resolveRunnerStorage, k as resolveShardLockDisableAdvisory, A as shardLockLogFields, F as workflowEngineLayer };

package/dist/clusterMigrationGate.js

@@ -0,0 +1,61 @@
+import { webCryptoLayer as e } from "./webCrypto.js";
+import { Effect as t, Layer as n } from "effect";
+import { SqlClient as r } from "effect/sql";
+import { ShardingConfig as i, SqlMessageStorage as a, SqlRunnerStorage as o } from "effect/cluster";
+import { applyUndeclaredCharsetWork as s, fnv1a64 as c, readUndeclaredCharsetColumns as l, resolveMigrationLockSchema as u, sqlPoolCapacity as d, undeclaredCharsetWork as f } from "@voltro/database/sql";
+//#region src/clusterMigrationGate.ts
+var p = 6322741009312601n, m = () => {
+	let e = u();
+	return e === void 0 ? p : c(`voltro_cluster_migration_lock:${e}`) & 9223372036854775807n;
+}, h = () => {
+	let e = `voltro_cluster_migration_${p.toString(36)}`, t = u();
+	if (t === void 0) return e;
+	let n = `${e}_${t}`;
+	return /^[A-Za-z0-9_]+$/.test(t) && n.length <= 64 ? n : `${e}_h${c(`voltro_cluster_migration_lock:${t}`).toString(16)}`;
+}, g = () => {
+	let e = Number(process.env.VOLTRO_CLUSTER_MIGRATION_LOCK_TIMEOUT_MS);
+	return Number.isFinite(e) && e > 0 ? e : 6e4;
+}, _ = 250, v = (e) => e[0]?.[0], y = (e, n) => e.onDialectOrElse({
+	sqlite: () => t.void,
+	mysql: () => n.executeValues(`SELECT GET_LOCK('${h()}', ${Math.ceil(g() / 1e3)})`, []).pipe(t.flatMap((e) => v(e) === 1 || v(e) === "1" ? t.void : t.die(/* @__PURE__ */ Error(`could not acquire cluster migration lock '${h()}' (GET_LOCK timed out)`)))),
+	mssql: () => n.executeValues(`DECLARE @res int; EXEC @res = sp_getapplock @Resource = '${h()}', @LockMode = 'Exclusive', @LockOwner = 'Session', @LockTimeout = ${g()}; SELECT @res`, []).pipe(t.flatMap((e) => Number(v(e) ?? -999) >= 0 ? t.void : t.die(/* @__PURE__ */ Error(`could not acquire cluster migration lock '${h()}' (sp_getapplock ${String(v(e))})`)))),
+	orElse: () => {
+		let e = m().toString(), r = Date.now(), i = n.executeValues(`SELECT pg_try_advisory_lock(${e}::bigint)`, []).pipe(t.flatMap((e) => {
+			let n = v(e);
+			return n === !0 || n === "t" || n === 1 ? t.void : Date.now() - r >= g() ? t.die(/* @__PURE__ */ Error(`could not acquire the cluster migration advisory lock within ${Math.round(g() / 1e3)}s. Another runner is migrating the cluster schema, or a prior boot wedged while holding it. (Raise VOLTRO_CLUSTER_MIGRATION_LOCK_TIMEOUT_MS, or inspect pg_locks for the holder.)`)) : t.sleep(`${_} millis`).pipe(t.andThen(i));
+		}));
+		return i;
+	}
+}), b = (e, n) => e.onDialectOrElse({
+	sqlite: () => t.void,
+	mysql: () => t.asVoid(n.executeValues(`SELECT RELEASE_LOCK('${h()}')`, [])),
+	mssql: () => t.asVoid(n.executeValues(`EXEC sp_releaseapplock @Resource = '${h()}', @LockOwner = 'Session'`, [])),
+	orElse: () => t.asVoid(n.executeValues(`SELECT pg_advisory_unlock(${m().toString()}::bigint)`, []))
+}), x = (e, n) => e.onDialectOrElse({
+	sqlite: () => !0,
+	orElse: () => !1
+}) ? n : t.scoped(t.gen(function* () {
+	let r = yield* e.reserve;
+	return yield* t.acquireRelease(y(e, r), () => b(e, r).pipe(t.orDie)), yield* n;
+})), S = (e) => e === "sql" ? ["cluster_messages", "cluster_runners"] : ["cluster_messages"], C = (e, n) => e.onDialectOrElse({
+	pg: () => e`SELECT (to_regclass(${n}) IS NOT NULL) AS present`,
+	mysql: () => e`SELECT (COUNT(*) > 0) AS present FROM information_schema.tables WHERE table_schema = DATABASE() AND table_name = ${n}`,
+	mssql: () => e`SELECT CASE WHEN OBJECT_ID(${n}, 'U') IS NULL THEN 0 ELSE 1 END AS present`,
+	orElse: () => e`SELECT (COUNT(*) > 0) AS present FROM sqlite_master WHERE type = 'table' AND name = ${n}`
+}).pipe(t.map((e) => {
+	let t = e[0]?.present;
+	return t === !0 || t === 1 || t === "1" || t === "t" || typeof t == "bigint" && t === 1n;
+}), t.catch(() => t.succeed(!1))), w = (e, n) => t.forEach(S(n), (t) => C(e, t), { concurrency: 1 }).pipe(t.map((e) => e.every(Boolean))), T = (a, o) => t.scoped(t.asVoid(n.build(o.pipe(n.provide(i.layerDefaults), n.provide(n.succeed(r.SqlClient, a)), n.provide(e))))), E = (e, n) => T(e, a.layer).pipe(t.andThen(n === "sql" ? T(e, o.layer) : t.void)), D = (e) => t.map(e.unsafe("SELECT table_name AS t FROM information_schema.tables WHERE table_schema = DATABASE() AND table_name LIKE 'cluster\\_%'"), (e) => new Set(e.map((e) => String(e.t ?? e.T)))), O = (e, n, r) => r === void 0 || !e.onDialectOrElse({
+	mysql: () => !0,
+	orElse: () => !1
+}) ? E(e, n) : t.gen(function* () {
+	let t = yield* D(e);
+	yield* E(e, n);
+	let i = new Set([...yield* D(e)].filter((e) => !t.has(e)));
+	i.size !== 0 && (yield* s(e, f(yield* l(e), /* @__PURE__ */ new Set(), r, i)));
+}), k = (e) => e === "sql" ? 3 : 2, A = (e) => n.effect(r.SqlClient, t.gen(function* () {
+	let n = yield* r.SqlClient, i = k(e.runnerStorage), a = d(n);
+	return a !== void 0 && a < i ? yield* t.die(/* @__PURE__ */ Error(`the workflow runtime needs ${i} database connections at once and the pool has ${a}: ` + (e.runnerStorage === "sql" ? "on a first boot it holds its migration lock and its runner lock while the migration runs on a third, and once running it keeps the runner lock and needs one more to work. " : "on a first boot it holds its migration lock while the migration runs on a second. ") + `Set DB_MAX_CONNECTIONS (or PG_MAX_CONNECTIONS) to at least ${i}` + (e.runnerStorage === "sql" ? ", or VOLTRO_WORKFLOW_RUNNER_STORAGE=memory for a single-process runtime that needs 2 (no cross-pod shard handoff)." : "."))) : ((yield* w(n, e.runnerStorage)) || (yield* x(n, O(n, e.runnerStorage, e.charset))), n);
+}));
+//#endregion
+export { k as clusterConnectionFloor, A as clusterMigrationGateLayer, C as tableExists, x as withClusterMigrationLock };

package/dist/clusterRunnerLayer.js

@@ -0,0 +1,21 @@
+import { webCryptoLayer as e } from "./webCrypto.js";
+import { guardRunnerStorage as t } from "./runnerStorageGuard.js";
+import { Effect as n, Layer as r, Semaphore as i } from "effect";
+import { SqlClient as a } from "effect/sql";
+import { RunnerHealth as o, RunnerStorage as s, Runners as c, Sharding as l, ShardingConfig as u, SqlMessageStorage as d, SqlRunnerStorage as f } from "effect/cluster";
+//#region src/clusterRunnerLayer.ts
+var p = (e) => n.map(i.make(1), (t) => ({
+	...e,
+	release: (r, i) => t.withPermits(1)(n.suspend(() => e.release(r, i)))
+})), m = r.unwrap(n.map(a.SqlClient, (e) => e.onDialectOrElse({
+	sqlite: () => r.effect(s.RunnerStorage, n.flatMap(s.RunnerStorage, p)).pipe(r.provide(f.layer)),
+	orElse: () => f.layer
+}))), h = (e) => r.effect(s.RunnerStorage, n.gen(function* () {
+	let n = yield* a.SqlClient, r = yield* s.RunnerStorage;
+	return yield* t(r, {
+		...e,
+		probe: n`SELECT 1`
+	});
+})).pipe(r.provide(r.orDie(m))), g = (t) => l.layer.pipe(r.provideMerge(c.layerNoop), r.provideMerge(d.layer), r.provide([t.runnerStorage === "memory" ? s.layerMemory : h(t.guard), o.layerNoop]), r.provide(u.layerFromEnv(t.shardingConfig)), r.provide(e));
+//#endregion
+export { g as clusterRunnerLayer, p as limitSqliteShardReleases };

package/dist/clusterTestSuite.d.ts

@@ -1,9 +1,9 @@
-import { ConfigError } from 'effect';
+import { Config } from 'effect';
 import { Layer } from 'effect';
 import { MysqlCharset } from '@voltro/database/sql';
-import { Sharding } from '@effect/cluster';
-import { SqlClient } from '@effect/sql';
-import { SqlError } from '@effect/sql';
+import { Sharding } from 'effect/cluster';
+import { SqlClient } from 'effect/sql';
+import { SqlError } from 'effect/sql';
 
 /**
  * Everything a dialect supplies to run the cluster-engine suite against
@@ -14,7 +14,7 @@ export declare interface ClusterEngineFixture {
     readonly name: string;
     /** The dialect's `SqlClient` layer (postgres/mysql/mariadb/mssql/sqlite).
      *  Typed via the engine's own option so `@voltro/testing` needs no
-     *  direct `@effect/sql` dependency. */
+     *  direct `effect/sql` dependency. */
     readonly sqlClientLayer: WorkflowEngineLayerOptions['sqlClientLayer'];
     /** `'sql'` for server dialects, `'memory'` for sqlite. */
     readonly runnerStorage: 'sql' | 'memory';
@@ -75,20 +75,20 @@ export declare const runClusterEngineSuite: (fixture: ClusterEngineFixture) => v
 export declare const runJournalEncryptionSuite: (fixture: ClusterEngineFixture) => void;
 
 /**
- * Runner-storage backend for `@effect/cluster.SingleRunner`.
+ * Runner-storage backend for `effect/cluster.SingleRunner`.
  *
  *   - `'sql'`     — durable across crashes + multi-instance scale-out.
  *                   Used for postgres / mysql / mariadb / mssql.
  *   - `'memory'`  — durable replay within the SAME process; no
  *                   horizontal scale-out. Used for `sqlite` because
- *                   `@effect/cluster.SqlRunnerStorage` has no SQLite
+ *                   `effect/cluster.SqlRunnerStorage` has no SQLite
  *                   lock-acquisition branch (`pg_advisory_lock` and
  *                   `GET_LOCK` aren't available there).
  */
 declare type RunnerStorage = 'sql' | 'memory';
 
 /**
- * How `@effect/cluster`'s `SqlRunnerStorage` coordinates shard ownership across
+ * How `effect/cluster`'s `SqlRunnerStorage` coordinates shard ownership across
  * a fleet of runners:
  *
  *   - `'advisory'` — session-scoped DB advisory locks (`GET_LOCK` on
@@ -129,7 +129,7 @@ declare interface WorkflowEngineLayerOptions {
      * runs through it for sqlite, the layer's resource lifecycle keeps
      * the cluster's accounting clean.
      */
-    readonly sqlClientLayer: Layer.Layer<SqlClient.SqlClient, ConfigError.ConfigError | SqlError.SqlError, never>;
+    readonly sqlClientLayer: Layer.Layer<SqlClient.SqlClient, Config.ConfigError | SqlError.SqlError, never>;
     /** SQL ownership for shared primaries; memory for local/embedded stores. */
     readonly runnerStorage: RunnerStorage;
     /**
@@ -161,7 +161,7 @@ declare interface WorkflowEngineLayerOptions {
      * lease-expiry problem, NOT a polling one: a dead runner keeps its shards
      * until its heartbeat goes stale, so the floor is the lease TTL.
      *
-     * - `failoverLeaseSeconds` (default 35, `@effect/cluster`
+     * - `failoverLeaseSeconds` (default 35, `effect/cluster`
      *   `shardLockExpiration`) — how long after a runner's last heartbeat its
      *   shards become claimable. Lower = faster failover on a hard crash, at the
      *   cost of false-positive reclaims if a healthy runner is paused longer than
@@ -178,7 +178,7 @@ declare interface WorkflowEngineLayerOptions {
     readonly failoverLeaseSeconds?: number;
     readonly failoverHeartbeatSeconds?: number;
     /**
-     * NEW-MESSAGE latency across replicas (default 10, `@effect/cluster`
+     * NEW-MESSAGE latency across replicas (default 10, `effect/cluster`
      * `entityMessagePollInterval`). DISTINCT from failover: this is how long a
      * freshly-triggered workflow can wait before the replica that OWNS its shard
      * picks it up. When the triggering replica owns the shard it's immediate