@akash-chowdhury-24/deployhub 2.0.43 → 2.0.44

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@akash-chowdhury-24/deployhub",
-  "version": "2.0.43",
+  "version": "2.0.44",
   "description": "Zero-configuration deployment and artifact manager",
   "type": "module",
   "main": "./src/cli/index.js",

package/src/deployment/providers/docker.js

@@ -218,7 +218,24 @@ export function createDockerProvider(config, envName, env = process.env) {
       await session.exec(ssh, cmds.pull, {
         timeoutMs: Math.max(session.defaultExecTimeoutMs, 300_000),
       });
-      await session.exec(ssh, cmds.run);
+      // `docker rm -f` can return before the name is fully released on a
+      // contended dockerd (`--rm` + stop). Re-rm immediately before run and
+      // retry name conflicts instead of failing a good deploy.
+      let lastRunErr;
+      for (let attempt = 1; attempt <= 3; attempt++) {
+        await session.exec(ssh, cmds.rm);
+        try {
+          await session.exec(ssh, cmds.run);
+          lastRunErr = null;
+          break;
+        } catch (err) {
+          lastRunErr = err;
+          const msg = err instanceof Error ? err.message : String(err);
+          if (!/already in use|Conflict/i.test(msg) || attempt === 3) throw err;
+          await new Promise((r) => setTimeout(r, 1000 * attempt));
+        }
+      }
+      if (lastRunErr) throw lastRunErr;
       const inspected = await session.execUnchecked(
         ssh,
         buildDockerInspectPortsCommand(containerName)

package/src/deployment/providers/ssh.js

@@ -80,7 +80,8 @@ export function createSshProvider(config, envName, env = process.env) {
   const { connect, exec, defaultExecTimeoutMs } = session;
   const startStopTimeoutMs = Math.min(
     defaultExecTimeoutMs,
-    Number(env.DEPLOYHUB_SSH_START_TIMEOUT_MS) || 60_000
+    Number(env.DEPLOYHUB_SSH_START_TIMEOUT_MS || process.env.DEPLOYHUB_SSH_START_TIMEOUT_MS) ||
+      60_000
   );
 
   /**

package/src/deployment/ssh-connection.js

@@ -28,8 +28,12 @@ export function createSshExecSession(opts) {
   const log = opts.log || { info() {}, error() {} };
 
   // Defense-in-depth: never let a stuck SSH channel hang CI indefinitely.
-  // Override with DEPLOYHUB_SSH_EXEC_TIMEOUT_MS (ms).
-  const defaultExecTimeoutMs = Number(env.DEPLOYHUB_SSH_EXEC_TIMEOUT_MS) || 120_000;
+  // Override with DEPLOYHUB_SSH_EXEC_TIMEOUT_MS (ms). Production default is 120s.
+  // Fall back to process.env so sparse credential overlays in tests/CI do not
+  // drop a process-level timeout override.
+  const defaultExecTimeoutMs =
+    Number(env.DEPLOYHUB_SSH_EXEC_TIMEOUT_MS || process.env.DEPLOYHUB_SSH_EXEC_TIMEOUT_MS) ||
+    120_000;
 
   async function connect() {
     if (!host || !user) {

package/src/utils/docker-image-deploy.js

@@ -109,15 +109,36 @@ export function createDockerImageDeployContext(config, env = process.env, log) {
     if (!hasRegistryCredentials()) return;
     const registry = registryUrl || 'https://index.docker.io/v1/';
     log.info('Logging in to container registry...');
-    await execa(
-      'docker',
-      ['login', registry, '-u', registryUser, '--password-stdin'],
-      {
-        input: registryToken,
-        stdio: ['pipe', 'inherit', 'inherit'],
-        env: getDockerEnv(),
+    const attempts = 3;
+    const loopback = /^(127\.0\.0\.1|localhost|\[::1\])/i.test(String(registry));
+    let lastErr;
+    for (let i = 1; i <= attempts; i++) {
+      try {
+        await execa(
+          'docker',
+          ['login', registry, '-u', registryUser, '--password-stdin'],
+          {
+            input: registryToken,
+            stdio: ['pipe', 'pipe', 'pipe'],
+            env: getDockerEnv(),
+            timeout: 60_000,
+          }
+        );
+        return;
+      } catch (err) {
+        lastErr = err;
+        const execErr = /** @type {{ message?: string, stderr?: string, stdout?: string }} */ (err);
+        const combined = `${execErr?.message || ''} ${execErr?.stderr || ''} ${execErr?.stdout || ''}`;
+        const transient =
+          /timeout|deadline exceeded|connection refused|tls handshake|i\/o timeout|eof|reset by peer|502|503/i.test(
+            combined
+          );
+        if (i === attempts || (!transient && !loopback)) throw err;
+        log.warn(`docker login attempt ${i}/${attempts} failed, retrying...`);
+        await new Promise((r) => setTimeout(r, 2000 * i));
       }
-    );
+    }
+    throw lastErr;
   }
 
   /**