fix: retry transient organization verification failures in CI

README.md

@@ -111,7 +111,7 @@ Gitea의 스테이징 설정을 `.gitea/workflows/deploy.yml`에서 Worker 런
 | `STAGING_ORG_CONTEXT_TENANT_IDS` | Variables | `ORG_CONTEXT_TENANT_IDS` |
 | `STAGING_ORG_CONTEXT_CACHE_TTL_SECONDS` | Variables, 선택 | `ORG_CONTEXT_CACHE_TTL_SECONDS` (기본 300) |
 
-조회는 `SCOPED`, `identifiers`를 사용합니다. 기준 테넌트는 UUID JSON 배열로 등록하며 UUID v4를 허용합니다. 각 테넌트를 최대 3개씩 동시 조회하고, 완전한 응답만 캐시합니다. 조회 타임아웃은 요청당 5초이며 일부 테넌트 실패·범위 불일치·구성원 누락은 검증 실패로 처리합니다.
+조회는 `SCOPED`, `identifiers`를 사용합니다. 기준 테넌트는 UUID JSON 배열로 등록하며 UUID v4를 허용합니다. 각 테넌트를 최대 3개씩 동시 조회하고, 완전한 응답만 캐시합니다. 페이지 조회 타임아웃은 요청당 5초입니다. 배포 전 `npm run org:verify`는 요청당 15초를 허용하고, 타임아웃·통신 실패·HTTP 429/502/503/504에 한해 2초 간격으로 최대 3회 시도합니다. 인증 실패·범위 불일치·구성원 누락은 재시도 없이 검증 실패로 처리합니다.
 
 기본 `ORG_CONTEXT_ENABLED=false`에서는 조직 API를 호출하지 않습니다. `true`로 배포하면 `npm run org:verify`가 실제 키로 조직 API를 먼저 검증하며, 실패 시 새 Worker 배포를 진행하지 않습니다. 성공 시에는 조직·사용자·소속 건수만 로그에 출력합니다. 전체 응답·전화번호·키 원문은 출력하지 않습니다.
 

package.json

@@ -6,7 +6,7 @@
     "deploy": "wrangler deploy",
     "deploy:production": "node scripts/deploy-production.mjs",
     "r2:upload": "bash scripts/upload-r2.sh",
-    "check": "node --check src/index.js && node --check src/org-context.js && node --check egbim/app.js && node --check scripts/production-settings.mjs && node --check scripts/deploy-production.mjs && node --check scripts/verify-production.mjs && bash -n scripts/upload-r2.sh",
+    "check": "node --check src/index.js && node --check src/org-context.js && node --check egbim/app.js && node --check scripts/verify-org-context.mjs && node --check scripts/production-settings.mjs && node --check scripts/deploy-production.mjs && node --check scripts/verify-production.mjs && bash -n scripts/upload-r2.sh",
     "test": "node --test --test-isolation=none tests/*.test.js",
     "org:verify": "node scripts/verify-org-context.mjs",
     "production:verify": "node scripts/verify-production.mjs"

scripts/verify-org-context.mjs

@@ -1,19 +1,39 @@
 import { getOrgDirectory } from '../src/org-context.js';
 import { deploymentOrgSettings } from './org-context-settings.mjs';
+import { setTimeout as wait } from 'node:timers/promises';
+import { pathToFileURL } from 'node:url';
 
-try {
-  const { vars, secrets } = deploymentOrgSettings(process.env);
+export async function verifyOrganization(env = process.env, { fetchImpl = fetch, pause = wait, log = console.log } = {}) {
+  const { vars, secrets } = deploymentOrgSettings(env);
   if (vars.ORG_CONTEXT_ENABLED !== 'true') {
-    console.log('Organization integration disabled; existing behavior retained.');
-  } else {
-    const directory = await getOrgDirectory({ ...vars, ...secrets });
-    console.log(JSON.stringify({
-      status: 'verified', rootCount: JSON.parse(vars.ORG_CONTEXT_TENANT_IDS).length,
-      tenantCount: directory.tenantCount, userCount: directory.byUser.size,
-      membershipCount: directory.membershipCount
-    }));
+    log('Organization integration disabled; existing behavior retained.');
+    return;
+  }
+  for (let attempt = 1; attempt <= 3; attempt++) {
+    try {
+      // CI can wait for a slow upstream without extending the page request timeout.
+      const directory = await getOrgDirectory({ ...vars, ...secrets }, { fetchImpl, timeoutMs: 15000 });
+      const summary = {
+        status: 'verified', rootCount: JSON.parse(vars.ORG_CONTEXT_TENANT_IDS).length,
+        tenantCount: directory.tenantCount, userCount: directory.byUser.size,
+        membershipCount: directory.membershipCount
+      };
+      log(JSON.stringify(summary));
+      return summary;
+    } catch (error) {
+      const retryable = /^org_context_(timeout|request_failed|http_(429|502|503|504))$/.test(error?.message || '');
+      if (!retryable || attempt === 3) throw error;
+      log(`Organization verification attempt ${attempt}/3 failed (${error.message}); retrying.`);
+      await pause(2000);
+    }
+  }
+}
+
+if (process.argv[1] && import.meta.url === pathToFileURL(process.argv[1]).href) {
+  try {
+    await verifyOrganization();
+  } catch (error) {
+    console.error(/^org_context_[a-z0-9_]+$/.test(error?.message || '') ? error.message : 'Organization integration configuration or verification failed.');
+    process.exitCode = 1;
   }
-} catch (error) {
-  console.error(/^org_context_[a-z0-9_]+$/.test(error?.message || '') ? error.message : 'Organization integration configuration or verification failed.');
-  process.exitCode = 1;
 }

src/org-context.js

@@ -120,9 +120,10 @@ export function resolveUserOrganization(directory, user) {
   return selected ? { status: 'resolved', ...selected } : { status: 'ambiguous_membership' };
 }
 
-export async function getOrgDirectory(env, { fetchImpl = fetch, now = Date.now } = {}) {
+export async function getOrgDirectory(env, { fetchImpl = fetch, now = Date.now, timeoutMs = 5000 } = {}) {
   const config = orgContextConfig(env);
   if (!config) return null;
+  if (!Number.isFinite(timeoutMs) || timeoutMs <= 0 || timeoutMs > 60000) throw new Error('org_context_invalid_timeout');
   // Include credential rotation without storing the secret in cache keys or diagnostic output.
   const digest = new Uint8Array(await crypto.subtle.digest('SHA-256', new TextEncoder().encode(config.keySecret)));
   const fingerprint = Array.from(digest, (byte) => byte.toString(16).padStart(2, '0')).join('');
@@ -140,7 +141,7 @@ export async function getOrgDirectory(env, { fetchImpl = fetch, now = Date.now }
         url.searchParams.set('tenantId', config.roots[index]);
         url.searchParams.set('exposure', 'identifiers');
         const controller = new AbortController();
-        const timeout = setTimeout(() => controller.abort(), 5000);
+        const timeout = setTimeout(() => controller.abort(), timeoutMs);
         try {
           const response = await fetchImpl(url.toString(), {
             headers: { accept: 'application/json', 'X-Baron-Key-ID': config.keyId, 'X-Baron-Key-Secret': config.keySecret },

tests/org-verification.test.js

@@ -0,0 +1,86 @@
+import test from 'node:test';
+import assert from 'node:assert/strict';
+import { verifyOrganization } from '../scripts/verify-org-context.mjs';
+
+const ROOT = '038326b6-954a-48a7-a85f-efd83f62b82a';
+const USER = '01900000-0000-7000-8000-000000000001';
+let sequence = 0;
+const settings = () => ({
+  ORG_CONTEXT_ENABLED: 'true', ORG_CONTEXT_TENANT_IDS: JSON.stringify([ROOT]),
+  ORG_CONTEXT_KEY_ID: 'fixture-verification', ORG_CONTEXT_KEY_SECRET: 'fixture-secret-' + ++sequence
+});
+const directory = () => ({ scope: { tenantId: ROOT }, tree: { id: ROOT, name: '기준 조직', members: [{ id: USER }] } });
+const response = (body, status = 200) => Response.json(body, { status });
+
+for (const status of [429, 502, 503, 504]) {
+  test(`deployment verification recovers from HTTP ${status} and reports only verified directory counts`, async () => {
+    let calls = 0;
+    const logs = [], pauses = [];
+    const result = await verifyOrganization(settings(), {
+      fetchImpl: async () => ++calls === 1 ? response({}, status) : response(directory()),
+      pause: async (ms) => pauses.push(ms), log: (line) => logs.push(line)
+    });
+    assert.equal(calls, 2);
+    assert.deepEqual(pauses, [2000]);
+    assert.equal(result.status, 'verified');
+    assert.equal(result.userCount, 1);
+    assert.ok(!JSON.stringify(logs).includes('fixture-secret'));
+    assert.ok(!JSON.stringify(logs).includes(USER));
+  });
+}
+
+test('deployment verification allows 15 seconds per request and retries an actual abort', async (t) => {
+  const originalTimer = globalThis.setTimeout;
+  const timeouts = [];
+  t.mock.method(globalThis, 'setTimeout', (callback, delay) => {
+    timeouts.push(delay);
+    return originalTimer(callback, delay === 15000 ? 5 : delay);
+  });
+  let calls = 0;
+  const result = await verifyOrganization(settings(), {
+    fetchImpl: async (_url, { signal }) => {
+      if (++calls === 1) return new Promise((_resolve, reject) => {
+        signal.addEventListener('abort', () => reject(new DOMException('Aborted', 'AbortError')), { once: true });
+      });
+      return response(directory());
+    },
+    pause: async () => {}, log() {}
+  });
+  assert.equal(calls, 2);
+  assert.deepEqual(timeouts, [15000, 15000]);
+  assert.equal(result.status, 'verified');
+});
+
+test('persistent network failure still blocks deployment after three attempts without exposing upstream errors', async () => {
+  let calls = 0, pauses = 0;
+  const logs = [];
+  await assert.rejects(verifyOrganization(settings(), {
+    fetchImpl: async () => { calls++; throw new Error('private-network-detail'); },
+    pause: async () => { pauses++; }, log: (line) => logs.push(line)
+  }), /org_context_request_failed/);
+  assert.equal(calls, 3);
+  assert.equal(pauses, 2);
+  assert.ok(!JSON.stringify(logs).includes('private-network-detail'));
+});
+
+for (const [name, body, status, error] of [
+  ['unauthorized', {}, 401, 'http_401'],
+  ['forbidden', {}, 403, 'http_403'],
+  ['wrong scope', { scope: { tenantId: USER } }, 200, 'scope_mismatch'],
+  ['missing members', { scope: { tenantId: ROOT }, tree: { id: ROOT, name: 'Empty' } }, 200, 'members_missing']
+]) {
+  test(`deployment verification rejects ${name} immediately`, async () => {
+    let calls = 0;
+    await assert.rejects(verifyOrganization(settings(), {
+      fetchImpl: async () => { calls++; return response(body, status); },
+      pause: async () => assert.fail('must not retry a configuration or data error'), log() {}
+    }), new RegExp('org_context_' + error));
+    assert.equal(calls, 1);
+  });
+}
+
+test('disabled verification makes no upstream requests', async () => {
+  await verifyOrganization({ ORG_CONTEXT_ENABLED: 'false' }, {
+    fetchImpl: async () => assert.fail('disabled lookup'), log() {}
+  });
+});