feat(ql3): add strong cluster run management

2026-09-23 12:05:27 +08:00 · 2026-08-12 07:28:43 +08:00
parent e38b143dbb
commit c0ab62e64a
58 changed files with 3087 additions and 105 deletions
@@ -0,0 +1,141 @@
+'use strict';
+
+const assert = require('node:assert/strict');
+const { test } = require('node:test');
+
+const {
+  ClusterRunManagementTransportAuthenticationError,
+  ClusterRunManagementTransportRequestError,
+  createClusterRunManagementTransport,
+  normalizeClusterRunManagementCommand,
+} = require('@qinglong/cluster-admin/run-management-transport');
+
+const NOW = 1_000_000;
+
+function principal(overrides = {}) {
+  return {
+    subject: { type: 'user', id: 'operator-1' },
+    authenticationId: 'oidc:run-management-1',
+    authenticatedAtMs: 999_000,
+    expiresAtMs: 1_100_000,
+    assurance: 'multi_factor',
+    ...overrides,
+  };
+}
+
+function command(overrides = {}) {
+  return {
+    schemaVersion: 1,
+    operation: 'run.retry',
+    request: {
+      projectId: 'project-1',
+      sourceRunId: 'source-run-1',
+      requestId: 'request-1',
+      auditEventId: '019f9300-0000-4000-8000-000000000001',
+      failureAuditEventId: '019f9300-0000-4000-8000-000000000002',
+      body: {
+        schema: 'qinglong/run-manual-retry@v1',
+        mutationId: '019f9300-0000-4000-8000-000000000003',
+        expectedRunVersion: 7,
+        expectedRunStatus: 'failed',
+      },
+      ...overrides,
+    },
+  };
+}
+
+function retryResult() {
+  return {
+    status: 'accepted',
+    projectId: 'project-1',
+    sourceRunId: 'source-run-1',
+    sourceRunStatus: 'failed',
+    sourceRunVersion: 7,
+    runId: '019f9300-0000-4000-8000-000000000010',
+    retryOfRunId: 'source-run-1',
+    taskId: 'task-1',
+    taskRevision: `qltd:v1:1:${'a'.repeat(64)}`,
+    attemptId: '019f9300-0000-4000-8000-000000000011',
+    runStatus: 'queued',
+    runVersion: 2,
+    eventSequence: 2,
+    executorType: 'remote_worker',
+    executionRevisionDigest: 'b'.repeat(64),
+    createdAtMs: NOW,
+  };
+}
+
+test('routes one exact strong User retry and emits the shared response', async () => {
+  const calls = [];
+  const transport = createClusterRunManagementTransport({
+    now: () => NOW,
+    service: {
+      async retry(request) {
+        calls.push(request);
+        return retryResult();
+      },
+    },
+  });
+  const result = await transport.execute(command(), {
+    authenticate: async () => principal(),
+  });
+  assert.equal(calls.length, 1);
+  assert.equal(calls[0].mutationId, command().request.body.mutationId);
+  assert.equal(calls[0].principal.assurance, 'multi_factor');
+  assert.deepEqual(result, {
+    schemaVersion: 1,
+    operation: 'run.retry',
+    retry: {
+      schema: 'qinglong/run-manual-retry@v1',
+      ...retryResult(),
+    },
+  });
+});
+
+test('rejects weak or non-User identity before service authority', async () => {
+  let called = false;
+  const transport = createClusterRunManagementTransport({
+    now: () => NOW,
+    service: {
+      async retry() {
+        called = true;
+        return retryResult();
+      },
+    },
+  });
+  await assert.rejects(
+    transport.execute(command(), {
+      authenticate: async () => principal({ assurance: 'single_factor' }),
+    }),
+    ClusterRunManagementTransportAuthenticationError,
+  );
+  await assert.rejects(
+    transport.execute(command(), {
+      authenticate: async () =>
+        principal({ subject: { type: 'agent', id: 'agent-1' } }),
+    }),
+    ClusterRunManagementTransportAuthenticationError,
+  );
+  assert.equal(called, false);
+});
+
+test('rejects widened commands and ambiguous audit identity', () => {
+  assert.throws(
+    () => normalizeClusterRunManagementCommand({ ...command(), principal: principal() }),
+    ClusterRunManagementTransportRequestError,
+  );
+  assert.throws(
+    () =>
+      normalizeClusterRunManagementCommand(
+        command({ failureAuditEventId: command().request.auditEventId }),
+      ),
+    ClusterRunManagementTransportRequestError,
+  );
+  assert.throws(
+    () =>
+      normalizeClusterRunManagementCommand(
+        command({ body: { ...command().request.body, expectedRunStatus: 'lost' } }),
+      ),
+    ClusterRunManagementTransportRequestError,
+  );
+});