Add retry mechanisms for network operations to fix Connect Timeout Error

Co-authored-by: eifinger <1481961+eifinger@users.noreply.github.com>

src/download/checksum/update-known-checksums.ts

@@ -1,5 +1,11 @@
 import { promises as fs } from "node:fs";
 import * as tc from "@actions/tool-cache";
+import {
+  isRetryableError,
+  NonRetryableError,
+  RetryableError,
+  withRetry,
+} from "../../utils/retry";
 import { KNOWN_CHECKSUMS } from "./known-checksums";
 export async function updateChecksums(
   filePath: string,
@@ -60,6 +66,27 @@ async function getOrDownloadChecksum(
 }
 
 async function downloadAssetContent(downloadUrl: string): Promise<string> {
+  return await withRetry(
+    async () => {
+      try {
         const downloadPath = await tc.downloadTool(downloadUrl);
         return await fs.readFile(downloadPath, "utf8");
+      } catch (error) {
+        const err = error as Error;
+        if (isRetryableError(err)) {
+          throw new RetryableError(
+            `Failed to download checksum file: ${err.message}`,
+            err,
+          );
+        } else {
+          throw new NonRetryableError(
+            `Failed to download checksum file: ${err.message}`,
+            err,
+          );
+        }
+      }
+    },
+    { maxRetries: 3, timeoutMs: 30000 },
+    "download checksum file",
+  );
 }

src/download/download-version.ts

@@ -9,6 +9,12 @@ import * as pep440 from "@renovatebot/pep440";
 import * as semver from "semver";
 import { OWNER, REPO, TOOL_CACHE_NAME } from "../utils/constants";
 import type { Architecture, Platform } from "../utils/platforms";
+import {
+  isRetryableError,
+  NonRetryableError,
+  RetryableError,
+  withRetry,
+} from "../utils/retry";
 import { validateChecksum } from "./checksum/checksum";
 
 const PaginatingOctokit = Octokit.plugin(paginateRest, restEndpointMethods);
@@ -43,11 +49,29 @@ export async function downloadVersion(
   const downloadUrl = constructDownloadUrl(version, platform, arch);
   core.debug(`Downloading ruff from "${downloadUrl}" ...`);
 
-  const downloadPath = await tc.downloadTool(
+  const downloadPath = await withRetry(
-    downloadUrl,
+    async () => {
-    undefined,
+      try {
-    githubToken,
+        return await tc.downloadTool(downloadUrl, undefined, githubToken);
+      } catch (error) {
+        const err = error as Error;
+        if (isRetryableError(err)) {
+          throw new RetryableError(
+            `Failed to download ruff binary: ${err.message}`,
+            err,
           );
+        } else {
+          throw new NonRetryableError(
+            `Failed to download ruff binary: ${err.message}`,
+            err,
+          );
+        }
+      }
+    },
+    { maxRetries: 3, timeoutMs: 60000 }, // 60 second timeout for downloads
+    "download ruff binary",
+  );
+
   core.debug(`Downloaded ruff to "${downloadPath}"`);
   await validateChecksum(checkSum, downloadPath, arch, platform, version);
 
@@ -134,30 +158,70 @@ export async function resolveVersion(
 }
 
 async function getAvailableVersions(githubToken: string): Promise<string[]> {
+  return await withRetry(
+    async () => {
       try {
         const octokit = new PaginatingOctokit({
           auth: githubToken,
         });
         return await getReleaseTagNames(octokit);
       } catch (err) {
-    if ((err as Error).message.includes("Bad credentials")) {
+        const error = err as Error;
+        if (error.message.includes("Bad credentials")) {
           core.info(
             "No (valid) GitHub token provided. Falling back to anonymous. Requests might be rate limited.",
           );
           const octokit = new PaginatingOctokit();
           return await getReleaseTagNames(octokit);
         }
-    throw err;
+
+        if (isRetryableError(error)) {
+          throw new RetryableError(
+            `Failed to get available versions: ${error.message}`,
+            error,
+          );
+        } else {
+          throw new NonRetryableError(
+            `Failed to get available versions: ${error.message}`,
+            error,
+          );
         }
       }
+    },
+    { maxRetries: 3, timeoutMs: 30000 }, // 30 second timeout for API calls
+    "get available versions",
+  );
+}
 
 async function getReleaseTagNames(
   octokit: InstanceType<typeof PaginatingOctokit>,
 ): Promise<string[]> {
-  const response = await octokit.paginate(octokit.rest.repos.listReleases, {
+  const response = await withRetry(
+    async () => {
+      try {
+        return await octokit.paginate(octokit.rest.repos.listReleases, {
           owner: OWNER,
           repo: REPO,
         });
+      } catch (error) {
+        const err = error as Error;
+        if (isRetryableError(err)) {
+          throw new RetryableError(
+            `Failed to list GitHub releases: ${err.message}`,
+            err,
+          );
+        } else {
+          throw new NonRetryableError(
+            `Failed to list GitHub releases: ${err.message}`,
+            err,
+          );
+        }
+      }
+    },
+    { maxRetries: 3, timeoutMs: 30000 },
+    "list GitHub releases",
+  );
+
   const releaseTagNames = response.map((release) => release.tag_name);
   if (releaseTagNames.length === 0) {
     throw Error(
@@ -168,6 +232,8 @@ async function getReleaseTagNames(
 }
 
 async function getLatestVersion(githubToken: string) {
+  return await withRetry(
+    async () => {
       const octokit = new PaginatingOctokit({
         auth: githubToken,
       });
@@ -176,7 +242,8 @@ async function getLatestVersion(githubToken: string) {
       try {
         latestRelease = await getLatestRelease(octokit);
       } catch (err) {
-    if ((err as Error).message.includes("Bad credentials")) {
+        const error = err as Error;
+        if (error.message.includes("Bad credentials")) {
           core.info(
             "No (valid) GitHub token provided. Falling back to anonymous. Requests might be rate limited.",
           );
@@ -186,7 +253,18 @@ async function getLatestVersion(githubToken: string) {
           core.error(
             "Github API request failed while getting latest release. Check the GitHub status page for outages. Try again later.",
           );
-      throw err;
+
+          if (isRetryableError(error)) {
+            throw new RetryableError(
+              `Failed to get latest version: ${error.message}`,
+              error,
+            );
+          } else {
+            throw new NonRetryableError(
+              `Failed to get latest version: ${error.message}`,
+              error,
+            );
+          }
         }
       }
 
@@ -194,16 +272,42 @@ async function getLatestVersion(githubToken: string) {
         throw new Error("Could not determine latest release.");
       }
       return latestRelease.tag_name;
+    },
+    { maxRetries: 3, timeoutMs: 30000 },
+    "get latest version",
+  );
 }
 
 async function getLatestRelease(
   octokit: InstanceType<typeof PaginatingOctokit>,
 ) {
-  const { data: latestRelease } = await octokit.rest.repos.getLatestRelease({
+  return await withRetry(
+    async () => {
+      try {
+        const { data: latestRelease } =
+          await octokit.rest.repos.getLatestRelease({
             owner: OWNER,
             repo: REPO,
           });
         return latestRelease;
+      } catch (error) {
+        const err = error as Error;
+        if (isRetryableError(err)) {
+          throw new RetryableError(
+            `Failed to get latest release: ${err.message}`,
+            err,
+          );
+        } else {
+          throw new NonRetryableError(
+            `Failed to get latest release: ${err.message}`,
+            err,
+          );
+        }
+      }
+    },
+    { maxRetries: 3, timeoutMs: 30000 },
+    "get latest release",
+  );
 }
 
 function maxSatisfying(

src/update-known-checksums.ts

@@ -5,6 +5,12 @@ import { restEndpointMethods } from "@octokit/plugin-rest-endpoint-methods";
 import * as semver from "semver";
 import { updateChecksums } from "./download/checksum/update-known-checksums";
 import { OWNER, REPO } from "./utils/constants";
+import {
+  isRetryableError,
+  NonRetryableError,
+  RetryableError,
+  withRetry,
+} from "./utils/retry";
 
 const PaginatingOctokit = Octokit.plugin(paginateRest, restEndpointMethods);
 
@@ -12,12 +18,33 @@ async function run(): Promise<void> {
   const checksumFilePath = process.argv.slice(2)[0];
   const github_token = process.argv.slice(2)[1];
 
+  const response = await withRetry(
+    async () => {
+      try {
         const octokit = new PaginatingOctokit({ auth: github_token });
-
+        return await octokit.paginate(octokit.rest.repos.listReleases, {
-  const response = await octokit.paginate(octokit.rest.repos.listReleases, {
           owner: OWNER,
           repo: REPO,
         });
+      } catch (error) {
+        const err = error as Error;
+        if (isRetryableError(err)) {
+          throw new RetryableError(
+            `Failed to list releases for checksum update: ${err.message}`,
+            err,
+          );
+        } else {
+          throw new NonRetryableError(
+            `Failed to list releases for checksum update: ${err.message}`,
+            err,
+          );
+        }
+      }
+    },
+    { maxRetries: 3, timeoutMs: 60000 },
+    "list releases for checksum update",
+  );
+
   const downloadUrls: string[] = response.flatMap((release) =>
     release.assets
       .filter((asset) => asset.name.endsWith(".sha256"))

src/utils/retry.ts

@@ -0,0 +1,176 @@
+import * as core from "@actions/core";
+
+export interface RetryOptions {
+  maxRetries: number;
+  initialDelayMs: number;
+  maxDelayMs: number;
+  backoffMultiplier: number;
+  timeoutMs?: number;
+}
+
+export const DEFAULT_RETRY_OPTIONS: RetryOptions = {
+  backoffMultiplier: 2,
+  initialDelayMs: 1000,
+  maxDelayMs: 10000,
+  maxRetries: 3,
+  timeoutMs: 30000, // 30 second timeout
+};
+
+export class RetryableError extends Error {
+  constructor(
+    message: string,
+    public readonly cause?: Error,
+  ) {
+    super(message);
+    this.name = "RetryableError";
+  }
+}
+
+export class NonRetryableError extends Error {
+  constructor(
+    message: string,
+    public readonly cause?: Error,
+  ) {
+    super(message);
+    this.name = "NonRetryableError";
+  }
+}
+
+export async function withRetry<T>(
+  operation: () => Promise<T>,
+  options: Partial<RetryOptions> = {},
+  operationName = "operation",
+): Promise<T> {
+  const opts = { ...DEFAULT_RETRY_OPTIONS, ...options };
+  let lastError: Error | undefined;
+
+  for (let attempt = 0; attempt <= opts.maxRetries; attempt++) {
+    try {
+      if (attempt > 0) {
+        const delay = Math.min(
+          opts.initialDelayMs * opts.backoffMultiplier ** (attempt - 1),
+          opts.maxDelayMs,
+        );
+        core.info(
+          `Retrying ${operationName} (attempt ${attempt + 1}/${opts.maxRetries + 1}) after ${delay}ms delay...`,
+        );
+        await sleep(delay);
+      }
+
+      // Wrap operation with timeout if specified
+      if (opts.timeoutMs) {
+        return await withTimeout(operation(), opts.timeoutMs, operationName);
+      } else {
+        return await operation();
+      }
+    } catch (error) {
+      lastError = error as Error;
+
+      // Don't retry on non-retryable errors
+      if (lastError instanceof NonRetryableError) {
+        core.debug(
+          `Non-retryable error in ${operationName}: ${lastError.message}`,
+        );
+        throw lastError.cause || lastError;
+      }
+
+      // Log the error for debugging
+      core.debug(
+        `Attempt ${attempt + 1} failed for ${operationName}: ${lastError.message}`,
+      );
+
+      // If this was the last attempt, throw the error
+      if (attempt === opts.maxRetries) {
+        core.error(
+          `${operationName} failed after ${opts.maxRetries + 1} attempts. Last error: ${lastError.message}`,
+        );
+        throw lastError;
+      }
+    }
+  }
+
+  throw lastError || new Error(`${operationName} failed for unknown reason`);
+}
+
+export async function withTimeout<T>(
+  promise: Promise<T>,
+  timeoutMs: number,
+  operationName = "operation",
+): Promise<T> {
+  const timeoutPromise = new Promise<never>((_, reject) => {
+    setTimeout(() => {
+      reject(
+        new RetryableError(`${operationName} timed out after ${timeoutMs}ms`),
+      );
+    }, timeoutMs);
+  });
+
+  return Promise.race([promise, timeoutPromise]);
+}
+
+function sleep(ms: number): Promise<void> {
+  return new Promise((resolve) => setTimeout(resolve, ms));
+}
+
+export function isRetryableError(error: Error): boolean {
+  const message = error.message.toLowerCase();
+
+  // Network-related errors that should be retried
+  const retryableMessages = [
+    "connect timeout",
+    "connection timeout",
+    "timeout",
+    "econnreset",
+    "econnrefused",
+    "enotfound",
+    "network error",
+    "request timeout",
+    "socket timeout",
+    "fetch failed",
+    "connect etimedout",
+  ];
+
+  // HTTP status codes that should be retried
+  const retryableStatusCodes = [408, 429, 500, 502, 503, 504];
+
+  // Check for retryable messages
+  if (retryableMessages.some((msg) => message.includes(msg))) {
+    return true;
+  }
+
+  // Check for HTTP status codes in error message
+  const statusMatch = message.match(/status.*?(\d{3})/);
+  if (statusMatch) {
+    const statusCode = parseInt(statusMatch[1], 10);
+    if (retryableStatusCodes.includes(statusCode)) {
+      return true;
+    }
+  }
+
+  return false;
+}
+
+export function wrapWithRetryLogic<T>(
+  operation: () => Promise<T>,
+  operationName: string,
+  options: Partial<RetryOptions> = {},
+): () => Promise<T> {
+  return async () => {
+    try {
+      return await withRetry(operation, options, operationName);
+    } catch (error) {
+      const err = error as Error;
+      if (isRetryableError(err)) {
+        throw new RetryableError(
+          `${operationName} failed: ${err.message}`,
+          err,
+        );
+      } else {
+        throw new NonRetryableError(
+          `${operationName} failed: ${err.message}`,
+          err,
+        );
+      }
+    }
+  };
+}