backend/src/services/deploy.ts

@@ -233,7 +233,7 @@ async function firstDeploy(
 
   try {
     await appendLog(previewId, `[PP] SSH connected. Waiting for bootstrap to complete...\n`);
-    await waitForBootstrap(sshSession, 600_000);
+    await waitForBootstrap(sshSession, 900_000);
     await appendLog(previewId, `[PP] Bootstrap complete. Starting setup...\n`);
 
     if (repoConfig.aptPackages.length > 0) {

backend/src/services/ec2.ts

@@ -115,7 +115,9 @@ export async function createPreviewSecurityGroup(ec2: EC2Client, groupName: stri
 }
 
 const BOOTSTRAP_SCRIPT = `#!/bin/bash
-set -e
+set -euo pipefail
+exec > >(tee -a /var/log/pp-bootstrap.log) 2>&1
+trap 'touch /var/lib/pp-bootstrap-failed' ERR
 apt-get update -y
 apt-get install -y curl git unzip build-essential
 
@@ -125,17 +127,6 @@ systemctl enable docker
 systemctl start docker
 apt-get install -y docker-compose-plugin
 
-# Allow root SSH with key auth (sshd_config may default to prohibit-password or no)
-sed -i 's/^#*PermitRootLogin.*/PermitRootLogin without-password/' /etc/ssh/sshd_config
-# Copy ubuntu's authorized_keys to root so PP can SSH as root
-mkdir -p /root/.ssh
-chmod 700 /root/.ssh
-if [ -f /home/ubuntu/.ssh/authorized_keys ]; then
-  cp /home/ubuntu/.ssh/authorized_keys /root/.ssh/authorized_keys
-  chmod 600 /root/.ssh/authorized_keys
-fi
-systemctl reload sshd || service ssh reload
-
 # NVM + Node LTS (installed as root, available to root SSH sessions)
 export HOME=/root
 curl -o- https://raw.githubusercontent.com/nvm-sh/nvm/v0.39.7/install.sh | bash
@@ -144,8 +135,18 @@ source "$NVM_DIR/nvm.sh"
 nvm install --lts
 nvm alias default lts/*
 
-# Signal that bootstrap is complete
+# Do not permit PP's root SSH session until every bootstrap step succeeded.
+# That prevents PP from seeing SSH as ready while cloud-init is still mutating
+# the host and turns the completion marker into a reliable readiness signal.
 touch /var/lib/pp-bootstrap-done
+sed -i 's/^#*PermitRootLogin.*/PermitRootLogin without-password/' /etc/ssh/sshd_config
+mkdir -p /root/.ssh
+chmod 700 /root/.ssh
+if [ -f /home/ubuntu/.ssh/authorized_keys ]; then
+  cp /home/ubuntu/.ssh/authorized_keys /root/.ssh/authorized_keys
+  chmod 600 /root/.ssh/authorized_keys
+fi
+systemctl reload sshd || service ssh reload
 `;
 
 export async function launchInstance(opts: {
@@ -179,12 +180,17 @@ export async function launchInstance(opts: {
 export async function waitForInstanceRunning(ec2: EC2Client, instanceId: string, maxWaitMs = 300_000): Promise<string> {
   const start = Date.now();
   while (Date.now() - start < maxWaitMs) {
-    const res = await ec2.send(new DescribeInstancesCommand({
-      InstanceIds: [instanceId],
-    }));
-    const inst = res.Reservations?.[0]?.Instances?.[0];
-    if (inst?.State?.Name === "running" && inst.PublicIpAddress) {
-      return inst.PublicIpAddress;
+    try {
+      const res = await ec2.send(new DescribeInstancesCommand({
+        InstanceIds: [instanceId],
+      }));
+      const inst = res.Reservations?.[0]?.Instances?.[0];
+      if (inst?.State?.Name === "running" && inst.PublicIpAddress) {
+        return inst.PublicIpAddress;
+      }
+    } catch (e: any) {
+      // AWS eventual consistency: instance may not be visible immediately after RunInstances
+      if (e.name !== "InvalidInstanceID.NotFound") throw e;
     }
     await sleep(5000);
   }

backend/src/services/ssh.ts

@@ -63,22 +63,32 @@ function tryConnect(host: string, privateKey: string, timeoutMs: number): Promis
       username: "root",
       privateKey,
       readyTimeout: timeoutMs,
+      keepaliveInterval: 10000,
+      keepaliveCountMax: 3,
       algorithms: {
-        serverHostKey: ["ssh-rsa", "ecdsa-sha2-nistp256", "ecdsa-sha2-nistp384", "ecdsa-sha2-nistp521"],
+        serverHostKey: ["ssh-rsa", "ecdsa-sha2-nistp256", "ecdsa-sha2-nistp384", "ecdsa-sha2-nistp521", "ssh-ed25519"],
       },
     });
   });
 }
 
+const EXEC_TIMEOUT_MS = 30_000;
+
 function execOnConn(conn: Client, command: string): Promise<{ stdout: string; stderr: string; code: number }> {
   return new Promise((resolve, reject) => {
+    const timer = setTimeout(() => reject(new Error(`SSH exec timed out: ${command.slice(0, 60)}`)), EXEC_TIMEOUT_MS);
+
     conn.exec(command, (err, stream) => {
-      if (err) return reject(err);
+      if (err) {
+        clearTimeout(timer);
+        return reject(err);
+      }
       let stdout = "";
       let stderr = "";
       stream.on("data", (d: Buffer) => { stdout += d.toString(); });
       stream.stderr.on("data", (d: Buffer) => { stderr += d.toString(); });
       stream.on("close", (code: number) => {
+        clearTimeout(timer);
         resolve({ stdout, stderr, code: code ?? 0 });
       });
     });
@@ -88,8 +98,12 @@ function execOnConn(conn: Client, command: string): Promise<{ stdout: string; st
 export async function waitForBootstrap(session: SshSession, maxWaitMs = 600_000): Promise<void> {
   const start = Date.now();
   while (Date.now() - start < maxWaitMs) {
-    const res = await session.exec("test -f /var/lib/pp-bootstrap-done && echo done || echo waiting");
-    if (res.stdout.trim() === "done") return;
+    try {
+      const res = await session.exec("test -f /var/lib/pp-bootstrap-done && echo done || echo waiting");
+      if (res.stdout.trim() === "done") return;
+    } catch {
+      // transient SSH exec error (e.g. timeout, channel reset) — retry
+    }
     await sleep(10_000);
   }
   throw new Error("EC2 bootstrap did not complete within timeout");