feat(ql3): expose bounded worker session observation

This commit is contained in:
whyour
2026-08-20 12:55:44 +08:00
parent 6265e31dce
commit 4a4fa85f51
24 changed files with 1437 additions and 28 deletions
@@ -348,5 +348,10 @@ export const postgresqlMainMigrationManifest: MigrationStreamManifest =
checksum:
'2fcac38386581189db63faacff325356f11c4529a8db9cef6be1a1ca706aaf10',
}),
Object.freeze({
id: 'pg-0069-worker-session-management-observation',
checksum:
'1191255575589abc2686b391827607abddb4edb78007245dbaaf45dc1c4e5e8b',
}),
]),
});
@@ -71,6 +71,7 @@ import { pg0065ApprovedActionManualRecoveryMigration } from '../approved-action/
import { pg0066CancellationDispatchMigration } from '../run/migrations/pg-0066-cancellation-dispatch';
import { pg0067CancellationDispatchManagementMigration } from '../run-management/pg-0067-cancellation-dispatch-management';
import { pg0068CancellationDispatchProjectKeysetMigration } from '../run-management/pg-0068-cancellation-dispatch-project-keyset';
import { pg0069WorkerSessionManagementObservationMigration } from '../remote-execution/pg-0069-worker-session-management-observation';
export const postgresqlMainMigrationStream: MigrationStreamDefinition<PostgresMigrationContext> =
Object.freeze({
@@ -147,5 +148,6 @@ export const postgresqlMainMigrationStream: MigrationStreamDefinition<PostgresMi
pg0066CancellationDispatchMigration,
pg0067CancellationDispatchManagementMigration,
pg0068CancellationDispatchProjectKeysetMigration,
pg0069WorkerSessionManagementObservationMigration,
]),
});
@@ -0,0 +1,17 @@
import { CAPABILITIES_V67 } from '../run-management/pg-0068-cancellation-dispatch-project-keyset';
import { definePostgresSqlMigration } from '../migrations/sqlMigration';
export const CAPABILITIES_V68 = CAPABILITIES_V67.replace(
'"worker_session":1}',
'"worker_session":1,"worker_session_observation":1}',
);
export const pg0069WorkerSessionManagementObservationMigration =
definePostgresSqlMigration({
id: 'pg-0069-worker-session-management-observation',
statements: [
`REVOKE ALL ON "ql3"."worker_sessions" FROM ql3_worker_credential_manager`,
`GRANT SELECT ON "ql3"."worker_sessions" TO ql3_worker_credential_manager`,
`DO $ql3$ BEGIN UPDATE "ql3"."schema_capabilities" SET contract_version = 68, migration_id = 'pg-0069-worker-session-management-observation', capabilities = '${CAPABILITIES_V68}'::jsonb, updated_at_ms = floor(extract(epoch FROM transaction_timestamp()) * 1000)::bigint WHERE contract_name = 'control-core' AND contract_version = 67 AND migration_id = 'pg-0068-cancellation-dispatch-project-keyset' AND capabilities = '${CAPABILITIES_V67}'::jsonb; IF NOT FOUND THEN RAISE EXCEPTION 'control-core capability is not at version 67' USING ERRCODE = 'check_violation'; END IF; END $ql3$`,
],
});
@@ -0,0 +1,188 @@
import type { PostgresPool } from '@qinglong/runtime-core';
import {
WORKER_SESSION_STATUSES,
assertWorkerCapabilitiesSnapshot,
assertWorkerConcurrency,
assertWorkerId,
assertWorkerSessionId,
type WorkerSessionRecord,
type WorkerSessionStatus,
} from '@qinglong/runtime-core/worker-session';
import {
MAX_WORKER_SESSION_OBSERVATION_PAGE_SIZE,
projectWorkerSessionObservation,
summarizeWorkerSessionObservation,
type WorkerSessionInspection,
type WorkerSessionObservationPage,
} from '@qinglong/runtime-core/worker-session-observation';
type Row = Record<string, unknown>;
const SELECT_COLUMNS = `
worker_id AS "workerId",
session_id AS "sessionId",
generation AS "generation",
status AS "status",
version AS "version",
capabilities_json AS "capabilitiesJson",
capabilities_hash AS "capabilitiesHash",
max_concurrent_runs AS "maxConcurrentRuns",
available_slots AS "availableSlots",
registered_at_ms AS "registeredAtMs",
last_heartbeat_at_ms AS "lastHeartbeatAtMs",
lease_expires_at_ms AS "leaseExpiresAtMs",
updated_at_ms AS "updatedAtMs"
`.trim();
function string(row: Row, key: string): string {
const value = row[key];
if (typeof value !== 'string' || value.length === 0) {
throw new TypeError(`PostgreSQL Worker observation ${key} is invalid`);
}
return value;
}
function integer(row: Row, key: string): number {
const value = row[key];
const normalized =
typeof value === 'string' && /^(0|[1-9]\d*)$/.test(value)
? Number(value)
: value;
if (!Number.isSafeInteger(normalized) || (normalized as number) < 0) {
throw new TypeError(`PostgreSQL Worker observation ${key} is invalid`);
}
return normalized as number;
}
function record(row: Row): Readonly<WorkerSessionRecord> {
const status = string(row, 'status');
if (!WORKER_SESSION_STATUSES.includes(status as WorkerSessionStatus)) {
throw new TypeError('PostgreSQL Worker observation status is invalid');
}
const worker: WorkerSessionRecord = Object.freeze({
workerId: string(row, 'workerId'),
sessionId: string(row, 'sessionId'),
generation: integer(row, 'generation'),
status: status as WorkerSessionStatus,
version: integer(row, 'version'),
capabilitiesJson: string(row, 'capabilitiesJson'),
capabilitiesHash: string(row, 'capabilitiesHash'),
maxConcurrentRuns: integer(row, 'maxConcurrentRuns'),
availableSlots: integer(row, 'availableSlots'),
registeredAtMs: integer(row, 'registeredAtMs'),
lastHeartbeatAtMs: integer(row, 'lastHeartbeatAtMs'),
leaseExpiresAtMs: integer(row, 'leaseExpiresAtMs'),
updatedAtMs: integer(row, 'updatedAtMs'),
});
assertWorkerId(worker.workerId);
assertWorkerSessionId(worker.sessionId);
assertWorkerCapabilitiesSnapshot(
worker.capabilitiesJson,
worker.capabilitiesHash,
);
assertWorkerConcurrency(worker.maxConcurrentRuns, worker.availableSlots);
return worker;
}
function observedAtMs(row: Row): number {
return integer(row, 'observedAtMs');
}
export class PostgresWorkerSessionObservationRepository {
constructor(private readonly pool: PostgresPool) {
if (!pool || typeof pool.query !== 'function') {
throw new TypeError('PostgreSQL Worker observation Pool is invalid');
}
}
async inspect(workerId: string): Promise<Readonly<WorkerSessionInspection>> {
assertWorkerId(workerId);
const result = await this.pool.query<Row>(
`WITH observation AS (
SELECT floor(extract(epoch FROM statement_timestamp()) * 1000)::bigint AS observed_at_ms
)
SELECT observation.observed_at_ms AS "observedAtMs", worker.*
FROM observation
LEFT JOIN LATERAL (
SELECT ${SELECT_COLUMNS}
FROM "ql3"."worker_sessions"
WHERE worker_id = $1
LIMIT 2
) AS worker ON TRUE`,
[workerId],
);
if (result.rows.length !== 1) {
throw new TypeError('PostgreSQL Worker inspection violated its bound');
}
const observed = observedAtMs(result.rows[0]!);
if (
result.rows[0]!.workerId !== null &&
string(result.rows[0]!, 'workerId') !== workerId
) {
throw new TypeError('PostgreSQL Worker inspection identity drifted');
}
return Object.freeze({
observedAtMs: observed,
worker:
result.rows[0]!.workerId === null
? null
: projectWorkerSessionObservation(record(result.rows[0]!), observed),
});
}
async list(
afterWorkerId: string | null,
): Promise<Readonly<WorkerSessionObservationPage>> {
if (afterWorkerId !== null) assertWorkerId(afterWorkerId);
const result = await this.pool.query<Row>(
`WITH observation AS (
SELECT floor(extract(epoch FROM statement_timestamp()) * 1000)::bigint AS observed_at_ms
), page AS (
SELECT ${SELECT_COLUMNS}
FROM "ql3"."worker_sessions"
WHERE ($1::varchar IS NULL OR worker_id > $1)
ORDER BY worker_id
LIMIT $2
)
SELECT observation.observed_at_ms AS "observedAtMs", page.*
FROM observation LEFT JOIN page ON TRUE
ORDER BY page."workerId"`,
[afterWorkerId, MAX_WORKER_SESSION_OBSERVATION_PAGE_SIZE + 1],
);
if (
result.rows.length < 1 ||
result.rows.length > MAX_WORKER_SESSION_OBSERVATION_PAGE_SIZE + 1
) {
throw new TypeError('PostgreSQL Worker observation page violated its bound');
}
const observed = observedAtMs(result.rows[0]!);
if (result.rows.some((row) => observedAtMs(row) !== observed)) {
throw new TypeError('PostgreSQL Worker observation clock drifted');
}
const observations =
result.rows[0]!.workerId === null
? []
: result.rows.map((row) =>
projectWorkerSessionObservation(record(row), observed),
);
for (let index = 0; index < observations.length; index += 1) {
const workerId = observations[index]!.workerId;
const previous = observations[index - 1]?.workerId ?? afterWorkerId;
if (previous !== null && workerId <= previous) {
throw new TypeError('PostgreSQL Worker observation page is unordered');
}
}
const page = observations.slice(
0,
MAX_WORKER_SESSION_OBSERVATION_PAGE_SIZE,
);
return Object.freeze({
observedAtMs: observed,
workers: Object.freeze(page.map(summarizeWorkerSessionObservation)),
nextCursor:
observations.length > MAX_WORKER_SESSION_OBSERVATION_PAGE_SIZE
? page.at(-1)!.workerId
: null,
});
}
}
@@ -21,8 +21,8 @@ export interface PostgresSchemaContractTrigger {
export interface PostgresSchemaContract {
readonly schema: 'ql3';
readonly contractName: 'control-core';
readonly contractVersion: 67;
readonly migrationId: 'pg-0068-cancellation-dispatch-project-keyset';
readonly contractVersion: 68;
readonly migrationId: 'pg-0069-worker-session-management-observation';
readonly minimumServerMajor: 16;
readonly maximumServerMajor: 18;
readonly capabilities: Readonly<{
@@ -92,6 +92,7 @@ export interface PostgresSchemaContract {
tool_invocation_artifact: 1;
trigger_definition: 1;
worker_session: 1;
worker_session_observation: 1;
worker_credential: 1;
worker_credential_delivery: 1;
worker_credential_execution_receipt: 1;
@@ -121,8 +122,8 @@ export const postgresqlControlSchemaContract: PostgresSchemaContract =
Object.freeze({
schema: 'ql3',
contractName: 'control-core',
contractVersion: 67,
migrationId: 'pg-0068-cancellation-dispatch-project-keyset',
contractVersion: 68,
migrationId: 'pg-0069-worker-session-management-observation',
minimumServerMajor: 16,
maximumServerMajor: 18,
capabilities: Object.freeze({
@@ -192,6 +193,7 @@ export const postgresqlControlSchemaContract: PostgresSchemaContract =
tool_invocation_artifact: 1,
trigger_definition: 1,
worker_session: 1,
worker_session_observation: 1,
worker_credential: 1,
worker_credential_delivery: 1,
worker_credential_execution_receipt: 1,
@@ -1509,7 +1509,8 @@ const REQUIRED_WORKER_CREDENTIAL_MANAGER_PRIVILEGES: RequiredPrivileges =
name === 'schema_migrations' ||
name === 'schema_capabilities' ||
name === 'projects' ||
name === 'project_role_bindings'
name === 'project_role_bindings' ||
name === 'worker_sessions'
? { ...NO_TABLE_PRIVILEGES, select: true }
: name === 'worker_credential_management_plans' ||
name === 'security_audit_events'
@@ -10,6 +10,7 @@ const OPERATIONS = [
'worker-credential.propose',
'worker-credential.decide',
'worker-credential.inspect',
'worker-session.observe',
] as const;
type Operation = (typeof OPERATIONS)[number];
type Row = Record<string, unknown>;
@@ -31,6 +32,7 @@ const DEFAULT_LIMITS: Readonly<Record<Operation, number>> = Object.freeze({
'worker-credential.propose': 30,
'worker-credential.decide': 60,
'worker-credential.inspect': 600,
'worker-session.observe': 600,
});
const ID = /^[A-Za-z0-9][A-Za-z0-9._:/-]{0,254}$/;
const PROJECT = /^[A-Za-z0-9][A-Za-z0-9._:-]{0,127}$/;
@@ -44,6 +44,7 @@ export { postgresqlMainMigrationManifest } from '../migration/migrationManifest'
export { PostgresApprovalRequestRepository } from '../approved-action/approvalRequestRepository';
export { PostgresProjectPolicyRepository } from '../security/projectPolicyRepository';
export { PostgresWorkerCredentialManagementPlanRepository } from './workerCredentialManagementPlanRepository';
export { PostgresWorkerSessionObservationRepository } from '../remote-execution/workerSessionObservationRepository';
export {
PostgresPluginPackageIdentityKeysetLedgerRepository as PostgresWorkerCredentialManagementIdentityKeysetLedgerRepository,
type ClusterManagementIdentityAuthority,
@@ -119,6 +119,7 @@ test('defines the immutable PostgreSQL capability and Run core stream', async ()
'pg-0066-cancellation-dispatch',
'pg-0067-cancellation-dispatch-management',
'pg-0068-cancellation-dispatch-project-keyset',
'pg-0069-worker-session-management-observation',
],
);
for (const migration of postgresqlMainMigrationStream.migrations) {
@@ -597,6 +598,11 @@ test('freezes every published PostgreSQL migration checksum', () => {
checksum:
'2fcac38386581189db63faacff325356f11c4529a8db9cef6be1a1ca706aaf10',
},
{
id: 'pg-0069-worker-session-management-observation',
checksum:
'1191255575589abc2686b391827607abddb4edb78007245dbaaf45dc1c4e5e8b',
},
];
assert.deepEqual(
postgresqlMainMigrationStream.migrations.map(({ id, checksum }) => ({
@@ -2421,3 +2427,36 @@ test('advances capability v67 with a Project-scoped blocked keyset', async () =>
/migration_id = 'pg-0067-cancellation-dispatch-management'/,
);
});
test('advances capability v68 with read-only Worker session observation', async () => {
const migration = migrationById(
'pg-0069-worker-session-management-observation',
);
const statements = [];
await migration.up({
async query(statement) {
statements.push(statement);
return { rows: [] };
},
});
const sql = statements.join('\n');
assert.match(
sql,
/REVOKE ALL ON "ql3"\."worker_sessions" FROM ql3_worker_credential_manager/,
);
assert.match(
sql,
/GRANT SELECT ON "ql3"\."worker_sessions" TO ql3_worker_credential_manager/,
);
assert.doesNotMatch(
sql,
/GRANT (?:INSERT|UPDATE|DELETE|TRUNCATE)[^;]+worker_sessions[^;]+ql3_worker_credential_manager/,
);
assert.match(sql, /contract_version = 68/);
assert.match(sql, /"worker_session_observation":1/);
assert.match(sql, /contract_version = 67/);
assert.match(
sql,
/migration_id = 'pg-0068-cancellation-dispatch-project-keyset'/,
);
});
@@ -561,6 +561,7 @@ function workerCredentialPrivileges(kind) {
? [
'worker_credential_management_quota_buckets',
'plugin_package_identity_keyset_ledger',
'worker_sessions',
]
: []),
...(manager
@@ -835,7 +836,7 @@ test('accepts the exact PostgreSQL control schema and least-privilege runtime ro
serverMajor: 16,
currentUser: 'ql3_runtime',
contractName: 'control-core',
contractVersion: 67,
contractVersion: 68,
migrationIds: [
'pg-0001-schema-capability',
'pg-0002-run-core',
@@ -905,6 +906,7 @@ test('accepts the exact PostgreSQL control schema and least-privilege runtime ro
'pg-0066-cancellation-dispatch',
'pg-0067-cancellation-dispatch-management',
'pg-0068-cancellation-dispatch-project-keyset',
'pg-0069-worker-session-management-observation',
],
});
});
@@ -935,10 +937,10 @@ test('accepts the exact schema and isolated least-privilege admin role', async (
}),
);
assert.equal(report.currentUser, 'ql3_admin');
assert.equal(report.contractVersion, 67);
assert.equal(report.contractVersion, 68);
assert.equal(
report.migrationIds.at(-1),
'pg-0068-cancellation-dispatch-project-keyset',
'pg-0069-worker-session-management-observation',
);
});
@@ -951,10 +953,10 @@ test('accepts the isolated least-privilege automation manager role', async () =>
}),
);
assert.equal(report.currentUser, 'ql3_automation_manager');
assert.equal(report.contractVersion, 67);
assert.equal(report.contractVersion, 68);
assert.equal(
report.migrationIds.at(-1),
'pg-0068-cancellation-dispatch-project-keyset',
'pg-0069-worker-session-management-observation',
);
const widened = automationManagerPrivileges();
@@ -983,10 +985,10 @@ test('accepts the isolated least-privilege human Approval manager role', async (
}),
);
assert.equal(report.currentUser, 'ql3_approval_manager');
assert.equal(report.contractVersion, 67);
assert.equal(report.contractVersion, 68);
assert.equal(
report.migrationIds.at(-1),
'pg-0068-cancellation-dispatch-project-keyset',
'pg-0069-worker-session-management-observation',
);
const widened = approvalManagerPrivileges();
@@ -1017,10 +1019,10 @@ test('accepts the isolated least-privilege Run manager role', async () => {
}),
);
assert.equal(report.currentUser, 'ql3_run_manager');
assert.equal(report.contractVersion, 67);
assert.equal(report.contractVersion, 68);
assert.equal(
report.migrationIds.at(-1),
'pg-0068-cancellation-dispatch-project-keyset',
'pg-0069-worker-session-management-observation',
);
const widened = runManagerPrivileges();
@@ -1181,10 +1183,10 @@ test('accepts the exact schema and isolated Worker ingress role', async () => {
}),
);
assert.equal(report.currentUser, 'ql3_worker_ingress');
assert.equal(report.contractVersion, 67);
assert.equal(report.contractVersion, 68);
assert.equal(
report.migrationIds.at(-1),
'pg-0068-cancellation-dispatch-project-keyset',
'pg-0069-worker-session-management-observation',
);
});
@@ -0,0 +1,119 @@
const assert = require('node:assert/strict');
const { test } = require('node:test');
const {
canonicalRemoteWorkerCapabilities,
} = require('@qinglong/runtime-core/remote-dispatch');
const {
PostgresWorkerSessionObservationRepository,
} = require('@qinglong/cluster-postgres/worker-credential-manager');
function row(index, overrides = {}) {
const canonical = canonicalRemoteWorkerCapabilities(
overrides.capabilities ?? {
architecture: index % 2 === 0 ? 'arm64' : 's390x',
executors: ['remote-worker'],
protocolVersion: '1.0.0',
supportTier: index % 2 === 0 ? 'tier1' : 'candidate',
runtimes: [{ name: 'node', version: '24.18.0' }],
},
);
return {
observedAtMs: 2_000,
workerId: `worker-${String(index).padStart(2, '0')}`,
sessionId: `018f0f5d-7b6a-7a11-8f4d-${String(index + 1).padStart(
12,
'0',
)}`,
generation: 1,
status: 'online',
version: 2,
capabilitiesJson: canonical.json,
capabilitiesHash: canonical.hash,
maxConcurrentRuns: 4,
availableSlots: 2,
registeredAtMs: 1_000,
lastHeartbeatAtMs: 1_900,
leaseExpiresAtMs: 3_000,
updatedAtMs: 1_900,
...Object.fromEntries(
Object.entries(overrides).filter(([key]) => key !== 'capabilities'),
),
};
}
test('inspects one exact Worker without exposing raw capability content', async () => {
const queries = [];
const repository = new PostgresWorkerSessionObservationRepository({
async query(text, values) {
queries.push({ text, values });
return { rows: [row(0)], rowCount: 1 };
},
});
const result = await repository.inspect('worker-00');
assert.equal(result.observedAtMs, 2_000);
assert.equal(result.worker.workerId, 'worker-00');
assert.equal(result.worker.compatibility, 'default_placement');
assert.deepEqual(result.worker.runtimes, [
{ name: 'node', version: '24.18.0' },
]);
assert.equal(Object.hasOwn(result.worker, 'capabilitiesJson'), false);
assert.deepEqual(queries[0].values, ['worker-00']);
assert.match(queries[0].text, /LIMIT 2/);
});
test('returns a masked absent inspection and a fixed sixteen-item keyset page', async () => {
let call = 0;
const repository = new PostgresWorkerSessionObservationRepository({
async query(_text, values) {
call += 1;
if (call === 1) {
return {
rows: [{ observedAtMs: 2_000, workerId: null }],
rowCount: 1,
};
}
assert.deepEqual(values, [null, 17]);
return {
rows: Array.from({ length: 17 }, (_, index) => row(index)),
rowCount: 17,
};
},
});
assert.deepEqual(await repository.inspect('missing-worker'), {
observedAtMs: 2_000,
worker: null,
});
const page = await repository.list(null);
assert.equal(page.workers.length, 16);
assert.equal(page.nextCursor, 'worker-15');
assert.equal(page.workers[1].compatibility, 'explicit_placement_required');
assert.equal(Object.hasOwn(page.workers[0], 'runtimes'), false);
assert.equal(Object.hasOwn(page.workers[0], 'declaredCapacity'), false);
});
test('fails closed on identity, ordering and database-clock drift', async () => {
const identity = new PostgresWorkerSessionObservationRepository({
async query() {
return { rows: [row(1)], rowCount: 1 };
},
});
await assert.rejects(identity.inspect('worker-00'), /identity drifted/);
const unordered = new PostgresWorkerSessionObservationRepository({
async query() {
return { rows: [row(1), row(0)], rowCount: 2 };
},
});
await assert.rejects(unordered.list(null), /page is unordered/);
const clock = new PostgresWorkerSessionObservationRepository({
async query() {
return {
rows: [row(0), row(1, { observedAtMs: 2_001 })],
rowCount: 2,
};
},
});
await assert.rejects(clock.list(null), /clock drifted/);
});