feat: container liveness sync + node-runner slim base

- Add syncStatus() to InstanceService: detects crashed/stopped containers,
  marks them ERROR with last log line as context
- Reconcile now syncs container status first (detect dead before counting)
- Add 30s periodic sync loop in main.ts
- Switch node-runner from alpine to slim (Debian) for npm compatibility
  (fixes home-assistant-mcp-server binary not found on Alpine)

Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>

deploy/Dockerfile.node-runner

@@ -1,7 +1,8 @@
 # Base container for npm-based MCP servers (STDIO transport).
 # mcpd uses this image to run `npx -y <packageName>` when a server
 # has packageName but no dockerImage.
-FROM node:20-alpine
+# Using slim (Debian) instead of alpine for better npm package compatibility.
+FROM node:20-slim
 
 WORKDIR /mcp
 

src/cli/src/commands/logs.ts

@@ -6,31 +6,65 @@ export interface LogsCommandDeps {
   log: (...args: unknown[]) => void;
 }
 
+interface InstanceInfo {
+  id: string;
+  status: string;
+  containerId: string | null;
+}
+
 /**
  * Resolve a name/ID to an instance ID.
  * Accepts: instance ID, server name, or server ID.
- * For servers, picks the first RUNNING instance.
+ * For servers with multiple replicas, picks by --instance index or first RUNNING.
  */
-async function resolveInstanceId(client: ApiClient, nameOrId: string): Promise<string> {
+async function resolveInstance(
+  client: ApiClient,
+  nameOrId: string,
+  instanceIndex?: number,
+): Promise<{ instanceId: string; serverName?: string; replicaInfo?: string }> {
   // Try as instance ID first
   try {
     await client.get(`/api/v1/instances/${nameOrId}`);
-    return nameOrId;
+    return { instanceId: nameOrId };
   } catch {
     // Not a valid instance ID
   }
 
-  // Try as server name → find its instances
+  // Try as server name/ID → find its instances
   const servers = await client.get<Array<{ id: string; name: string }>>('/api/v1/servers');
   const server = servers.find((s) => s.name === nameOrId || s.id === nameOrId);
-  if (server) {
-    const instances = await client.get<Array<{ id: string; status: string }>>(`/api/v1/instances?serverId=${server.id}`);
-    const running = instances.find((i) => i.status === 'RUNNING') ?? instances[0];
-    if (running) return running.id;
-    throw new Error(`No instances found for server '${nameOrId}'`);
+  if (!server) {
+    throw new Error(`Instance or server '${nameOrId}' not found`);
   }
 
-  throw new Error(`Instance or server '${nameOrId}' not found`);
+  const instances = await client.get<InstanceInfo[]>(`/api/v1/instances?serverId=${server.id}`);
+  if (instances.length === 0) {
+    throw new Error(`No instances found for server '${server.name}'`);
+  }
+
+  // Select by index or pick first running
+  let selected: InstanceInfo | undefined;
+  if (instanceIndex !== undefined) {
+    if (instanceIndex < 0 || instanceIndex >= instances.length) {
+      throw new Error(`Instance index ${instanceIndex} out of range (server '${server.name}' has ${instances.length} instance${instances.length > 1 ? 's' : ''})`);
+    }
+    selected = instances[instanceIndex];
+  } else {
+    selected = instances.find((i) => i.status === 'RUNNING') ?? instances[0];
+  }
+
+  if (!selected) {
+    throw new Error(`No instances found for server '${server.name}'`);
+  }
+
+  const result: { instanceId: string; serverName?: string; replicaInfo?: string } = {
+    instanceId: selected.id,
+    serverName: server.name,
+  };
+  if (instances.length > 1) {
+    result.replicaInfo = `instance ${instances.indexOf(selected) + 1}/${instances.length}`;
+  }
+  return result;
 }
 
 export function createLogsCommand(deps: LogsCommandDeps): Command {
@@ -40,8 +74,15 @@ export function createLogsCommand(deps: LogsCommandDeps): Command {
     .description('Get logs from an MCP server instance')
     .argument('<name>', 'Server name, server ID, or instance ID')
     .option('-t, --tail <lines>', 'Number of lines to show')
-    .action(async (nameOrId: string, opts: { tail?: string }) => {
-      const instanceId = await resolveInstanceId(client, nameOrId);
+    .option('-i, --instance <index>', 'Instance/replica index (0-based, for servers with multiple replicas)')
+    .action(async (nameOrId: string, opts: { tail?: string; instance?: string }) => {
+      const instanceIndex = opts.instance !== undefined ? parseInt(opts.instance, 10) : undefined;
+      const { instanceId, serverName, replicaInfo } = await resolveInstance(client, nameOrId, instanceIndex);
+
+      if (replicaInfo) {
+        process.stderr.write(`Showing logs for ${serverName} (${replicaInfo})\n`);
+      }
+
       let url = `/api/v1/instances/${instanceId}/logs`;
       if (opts.tail) {
         url += `?tail=${opts.tail}`;

src/cli/tests/commands/instances.test.ts

@@ -68,16 +68,79 @@ describe('logs command', () => {
     output = [];
   });
 
-  it('shows logs', async () => {
-    vi.mocked(client.get).mockResolvedValue({ stdout: 'hello world\n', stderr: '' });
+  it('shows logs by instance ID', async () => {
+    vi.mocked(client.get)
+      .mockResolvedValueOnce({ id: 'inst-1', status: 'RUNNING' } as never) // instance lookup
+      .mockResolvedValueOnce({ stdout: 'hello world\n', stderr: '' } as never); // logs
     const cmd = createLogsCommand({ client, log });
     await cmd.parseAsync(['inst-1'], { from: 'user' });
+    expect(client.get).toHaveBeenCalledWith('/api/v1/instances/inst-1');
     expect(client.get).toHaveBeenCalledWith('/api/v1/instances/inst-1/logs');
     expect(output.join('\n')).toContain('hello world');
   });
 
+  it('resolves server name to instance ID', async () => {
+    vi.mocked(client.get)
+      .mockRejectedValueOnce(new Error('not found')) // instance lookup fails
+      .mockResolvedValueOnce([{ id: 'srv-1', name: 'my-grafana' }] as never) // servers list
+      .mockResolvedValueOnce([{ id: 'inst-1', status: 'RUNNING', containerId: 'abc' }] as never) // instances for server
+      .mockResolvedValueOnce({ stdout: 'grafana logs\n', stderr: '' } as never); // logs
+    const cmd = createLogsCommand({ client, log });
+    await cmd.parseAsync(['my-grafana'], { from: 'user' });
+    expect(client.get).toHaveBeenCalledWith('/api/v1/instances/inst-1/logs');
+    expect(output.join('\n')).toContain('grafana logs');
+  });
+
+  it('picks RUNNING instance over others', async () => {
+    vi.mocked(client.get)
+      .mockRejectedValueOnce(new Error('not found'))
+      .mockResolvedValueOnce([{ id: 'srv-1', name: 'ha-mcp' }] as never)
+      .mockResolvedValueOnce([
+        { id: 'inst-err', status: 'ERROR', containerId: null },
+        { id: 'inst-ok', status: 'RUNNING', containerId: 'abc' },
+      ] as never)
+      .mockResolvedValueOnce({ stdout: 'running instance\n', stderr: '' } as never);
+    const cmd = createLogsCommand({ client, log });
+    await cmd.parseAsync(['ha-mcp'], { from: 'user' });
+    expect(client.get).toHaveBeenCalledWith('/api/v1/instances/inst-ok/logs');
+  });
+
+  it('selects specific replica with --instance', async () => {
+    vi.mocked(client.get)
+      .mockRejectedValueOnce(new Error('not found'))
+      .mockResolvedValueOnce([{ id: 'srv-1', name: 'ha-mcp' }] as never)
+      .mockResolvedValueOnce([
+        { id: 'inst-0', status: 'RUNNING', containerId: 'a' },
+        { id: 'inst-1', status: 'RUNNING', containerId: 'b' },
+      ] as never)
+      .mockResolvedValueOnce({ stdout: 'replica 1\n', stderr: '' } as never);
+    const cmd = createLogsCommand({ client, log });
+    await cmd.parseAsync(['ha-mcp', '-i', '1'], { from: 'user' });
+    expect(client.get).toHaveBeenCalledWith('/api/v1/instances/inst-1/logs');
+  });
+
+  it('throws on out-of-range --instance index', async () => {
+    vi.mocked(client.get)
+      .mockRejectedValueOnce(new Error('not found'))
+      .mockResolvedValueOnce([{ id: 'srv-1', name: 'ha-mcp' }] as never)
+      .mockResolvedValueOnce([{ id: 'inst-0', status: 'RUNNING' }] as never);
+    const cmd = createLogsCommand({ client, log });
+    await expect(cmd.parseAsync(['ha-mcp', '-i', '5'], { from: 'user' })).rejects.toThrow('out of range');
+  });
+
+  it('throws when server has no instances', async () => {
+    vi.mocked(client.get)
+      .mockRejectedValueOnce(new Error('not found'))
+      .mockResolvedValueOnce([{ id: 'srv-1', name: 'empty-srv' }] as never)
+      .mockResolvedValueOnce([] as never);
+    const cmd = createLogsCommand({ client, log });
+    await expect(cmd.parseAsync(['empty-srv'], { from: 'user' })).rejects.toThrow('No instances found');
+  });
+
   it('passes tail option', async () => {
-    vi.mocked(client.get).mockResolvedValue({ stdout: '', stderr: '' });
+    vi.mocked(client.get)
+      .mockResolvedValueOnce({ id: 'inst-1' } as never)
+      .mockResolvedValueOnce({ stdout: '', stderr: '' } as never);
     const cmd = createLogsCommand({ client, log });
     await cmd.parseAsync(['inst-1', '-t', '50'], { from: 'user' });
     expect(client.get).toHaveBeenCalledWith('/api/v1/instances/inst-1/logs?tail=50');

src/mcpd/src/main.ts

@@ -134,9 +134,22 @@ async function main(): Promise<void> {
   await app.listen({ port: config.port, host: config.host });
   app.log.info(`mcpd listening on ${config.host}:${config.port}`);
 
+  // Periodic container liveness sync — detect crashed containers
+  const SYNC_INTERVAL_MS = 30_000; // 30s
+  const syncTimer = setInterval(async () => {
+    try {
+      await instanceService.syncStatus();
+    } catch (err) {
+      app.log.error({ err }, 'Container status sync failed');
+    }
+  }, SYNC_INTERVAL_MS);
+
   // Graceful shutdown
   setupGracefulShutdown(app, {
-    disconnectDb: () => prisma.$disconnect(),
+    disconnectDb: async () => {
+      clearInterval(syncTimer);
+      await prisma.$disconnect();
+    },
   });
 }
 

src/mcpd/src/services/instance.service.ts

@@ -36,8 +36,41 @@ export class InstanceService {
     return instance;
   }
 
+  /**
+   * Sync instance statuses with actual container state.
+   * Detects crashed/stopped containers and marks them ERROR.
+   */
+  async syncStatus(): Promise<void> {
+    const instances = await this.instanceRepo.findAll();
+    for (const inst of instances) {
+      if ((inst.status === 'RUNNING' || inst.status === 'STARTING') && inst.containerId) {
+        try {
+          const info = await this.orchestrator.inspectContainer(inst.containerId);
+          if (info.state === 'stopped' || info.state === 'error') {
+            // Container died — get last logs for error context
+            let errorMsg = `Container ${info.state}`;
+            try {
+              const logs = await this.orchestrator.getContainerLogs(inst.containerId, { tail: 5 });
+              const lastLog = (logs.stdout || logs.stderr).trim().split('\n').pop();
+              if (lastLog) errorMsg = lastLog;
+            } catch { /* best-effort */ }
+            await this.instanceRepo.updateStatus(inst.id, 'ERROR', {
+              metadata: { error: errorMsg },
+            });
+          }
+        } catch {
+          // Container gone entirely
+          await this.instanceRepo.updateStatus(inst.id, 'ERROR', {
+            metadata: { error: 'Container not found' },
+          });
+        }
+      }
+    }
+  }
+
   /**
    * Reconcile instances for a server to match desired replica count.
+   * - Syncs container statuses first (detect crashed containers)
    * - If fewer running instances than replicas: start new ones
    * - If more running instances than replicas: remove excess (oldest first)
    */
@@ -45,6 +78,9 @@ export class InstanceService {
     const server = await this.serverRepo.findById(serverId);
     if (!server) throw new NotFoundError(`McpServer '${serverId}' not found`);
 
+    // Sync container statuses before counting active instances
+    await this.syncStatus();
+
     const instances = await this.instanceRepo.findAll(serverId);
     const active = instances.filter((i) => i.status === 'RUNNING' || i.status === 'STARTING');
     const desired = server.replicas;