ci: Add reusable retry script for CI commands (#26815)

Co-authored-by: Claude Opus 4.6 <noreply@anthropic.com>

.github/actions/docker-registry-login/action.yml

@@ -29,11 +29,13 @@ runs:
   steps:
     - name: Login to GitHub Container Registry
       if: inputs.login-ghcr == 'true'
-      uses: docker/login-action@5e57cd118135c172c3672efd75eb46360885c0ef # v3.6.0
-      with:
-        registry: ghcr.io
-        username: ${{ github.actor }}
-        password: ${{ github.token }}
+      shell: bash
+      env:
+        GHCR_TOKEN: ${{ github.token }}
+        GHCR_USER: ${{ github.actor }}
+      run: |
+        node .github/scripts/retry.mjs --attempts 3 --delay 10 \
+          'echo "$GHCR_TOKEN" | docker login ghcr.io -u "$GHCR_USER" --password-stdin'
 
     - name: Login to DockerHub
       if: inputs.login-dockerhub == 'true'

.github/scripts/retry.mjs

@@ -0,0 +1,60 @@
+#!/usr/bin/env node
+/**
+ * Retry a shell command with configurable attempts and delay.
+ *
+ * Usage: node retry.mjs [--attempts N] [--delay N] '<command>'
+ *
+ * Options:
+ *   --attempts N   Maximum number of attempts (default: 4)
+ *   --delay N      Seconds to wait between retries (default: 15)
+ *
+ * The command is executed via shell, so pipes and env-var expansion work.
+ * Exits 0 on first success, 1 if all attempts fail.
+ */
+import { execSync } from 'node:child_process';
+
+const args = process.argv.slice(2);
+
+function getFlag(name, defaultValue) {
+	const index = args.indexOf(`--${name}`);
+	if (index === -1 || !args[index + 1]) return defaultValue;
+	const value = parseInt(args[index + 1], 10);
+	if (Number.isNaN(value) || value <= 0) {
+		console.error(`Error: --${name} must be a positive integer`);
+		process.exit(1);
+	}
+	return value;
+}
+
+const attempts = getFlag('attempts', 4);
+const delay = getFlag('delay', 15);
+
+// Command is the last positional arg (skip flags and their values)
+const command = args
+	.filter((a, i) => {
+		if (a.startsWith('--')) return false;
+		if (i > 0 && args[i - 1].startsWith('--')) return false;
+		return true;
+	})
+	.pop();
+
+if (!command) {
+	console.error("Usage: node retry.mjs [--attempts N] [--delay N] '<command>'");
+	process.exit(1);
+}
+
+for (let i = 1; i <= attempts; i++) {
+	try {
+		execSync(command, { shell: true, stdio: 'inherit' });
+		process.exit(0);
+	} catch {
+		if (i < attempts) {
+			console.error(`Attempt ${i}/${attempts} failed, retrying in ${delay}s...`);
+			execSync(`sleep ${delay}`);
+		} else {
+			console.error(`Attempt ${i}/${attempts} failed, no more retries.`);
+		}
+	}
+}
+
+process.exit(1);

.github/workflows/security-trivy-scan-callable.yml

@@ -36,14 +36,11 @@ jobs:
             security/vex.openvex.json
             security/trivy.yaml
             security/trivy-ignore-policy.rego
+            .github/scripts/retry.mjs
           sparse-checkout-cone-mode: false
 
       - name: Pull Docker image with retry
-        run: |
-          for i in {1..4}; do
-            docker pull "${{ inputs.image_ref }}" && break
-            [ "$i" -lt 4 ] && echo "Retry $i failed, waiting..." && sleep 15
-          done
+        run: node .github/scripts/retry.mjs --attempts 4 --delay 15 'docker pull "${{ inputs.image_ref }}"'
 
       - name: Run Trivy vulnerability scanner
         uses: aquasecurity/trivy-action@e368e328979b113139d6f9068e03accaed98a518 # v0.34.1

packages/testing/containers/pull-test-images.ts

@@ -24,24 +24,42 @@ interface PullResult {
 	pulled: number;
 }
 
+const MAX_RETRIES = 3;
+const RETRY_DELAY_MS = 10_000;
+
+async function sleep(ms: number): Promise<void> {
+	return await new Promise((resolve) => setTimeout(resolve, ms));
+}
+
 async function pullImage(image: string): Promise<PullResult> {
 	const imageStart = Date.now();
-	try {
-		const { stdout, stderr } = await execAsync(`docker pull ${image}`);
-		const output = stdout + stderr;
 
-		// Count cached vs pulled layers
-		const cached = (output.match(/Already exists/g) ?? []).length;
-		const pulled = (output.match(/Pull complete/g) ?? []).length;
+	for (let attempt = 1; attempt <= MAX_RETRIES; attempt++) {
+		try {
+			const { stdout, stderr } = await execAsync(`docker pull ${image}`);
+			const output = stdout + stderr;
 
-		const duration = ((Date.now() - imageStart) / 1000).toFixed(1);
-		return { image, duration, success: true, cached, pulled };
-	} catch (error) {
-		const duration = ((Date.now() - imageStart) / 1000).toFixed(1);
-		const message = error instanceof Error ? error.message : String(error);
-		console.error(`   ⚠️  Pull failed for ${image}: ${message}`);
-		return { image, duration, success: false, cached: 0, pulled: 0 };
+			const cached = (output.match(/Already exists/g) ?? []).length;
+			const pulled = (output.match(/Pull complete/g) ?? []).length;
+
+			const duration = ((Date.now() - imageStart) / 1000).toFixed(1);
+			return { image, duration, success: true, cached, pulled };
+		} catch (error) {
+			const message = error instanceof Error ? error.message : String(error);
+			if (attempt < MAX_RETRIES) {
+				console.warn(
+					`   ⚠️  Pull attempt ${attempt}/${MAX_RETRIES} failed for ${image}: ${message}`,
+				);
+				console.warn(`   ⏳ Retrying in ${RETRY_DELAY_MS / 1000}s...`);
+				await sleep(RETRY_DELAY_MS);
+			} else {
+				console.error(`   ❌ Pull failed for ${image} after ${MAX_RETRIES} attempts: ${message}`);
+			}
+		}
 	}
+
+	const duration = ((Date.now() - imageStart) / 1000).toFixed(1);
+	return { image, duration, success: false, cached: 0, pulled: 0 };
 }
 
 function isValidImageKey(key: string): key is ImageKey {