fix(container): scope orphan reaper by install label so peers don't kill each other

Two installs on the same host could trash each other's containers: the
reaper used `docker ps --filter name=nanoclaw-`, a substring match that
picked up every install's containers. A crash-looping peer (e.g. a legacy
v1 plist respawning ~6k times) would call cleanupOrphans on every boot and
kill the healthy install's session containers within seconds of spawn.

- Stamp `--label nanoclaw-install=<slug>` onto every spawned container.
- cleanupOrphans filters by that label; healthy peers are left alone.
- Setup preflight enumerates `com.nanoclaw*` launchd plists / nanoclaw
  user systemd units, probes state/runs, and unloads any that are
  crash-looping (state != running AND runs > 10) before installing
  this install's service.

Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>

setup/peer-cleanup.ts

@@ -0,0 +1,186 @@
+/**
+ * Detect and clean up unhealthy NanoClaw peer services.
+ *
+ * Runs as a setup preflight before we install our own service. A crash-looping
+ * peer install (typically the legacy v1 `com.nanoclaw` plist) silently trashes
+ * this install's containers on every respawn because its `cleanupOrphans()`
+ * reaps anything matching `nanoclaw-`. We scope our reaper by label now, but
+ * we still need to stop the peer from killing us on its way down.
+ *
+ * A peer is "unhealthy" when:
+ *   - launchd: `state != running` AND `runs > UNHEALTHY_RUNS_THRESHOLD`
+ *   - systemd: unit is in `failed` state, OR `activating` with many restarts
+ *
+ * Healthy peers are left alone — multiple installs can coexist fine now that
+ * container-reaper is label-scoped.
+ */
+import { execFileSync } from 'child_process';
+import fs from 'fs';
+import os from 'os';
+import path from 'path';
+
+import { getLaunchdLabel, getSystemdUnit } from '../src/install-slug.js';
+import { log } from '../src/log.js';
+
+const UNHEALTHY_RUNS_THRESHOLD = 10;
+
+export interface PeerStatus {
+  label: string;
+  configPath: string;
+  state: string;
+  runs: number;
+  unhealthy: boolean;
+}
+
+export interface PeerCleanupResult {
+  checked: PeerStatus[];
+  unloaded: PeerStatus[];
+  failures: Array<{ label: string; err: string }>;
+}
+
+/**
+ * Scan for peer NanoClaw services and unload any that are crash-looping.
+ * Returns a summary suitable for emitStatus / setup-log reporting.
+ */
+export function cleanupUnhealthyPeers(projectRoot: string = process.cwd()): PeerCleanupResult {
+  const platform = os.platform();
+  if (platform === 'darwin') {
+    return cleanupLaunchdPeers(projectRoot);
+  }
+  if (platform === 'linux') {
+    return cleanupSystemdPeers(projectRoot);
+  }
+  return { checked: [], unloaded: [], failures: [] };
+}
+
+// ---- launchd (macOS) --------------------------------------------------------
+
+function cleanupLaunchdPeers(projectRoot: string): PeerCleanupResult {
+  const ownLabel = getLaunchdLabel(projectRoot);
+  const agentsDir = path.join(os.homedir(), 'Library', 'LaunchAgents');
+  const result: PeerCleanupResult = { checked: [], unloaded: [], failures: [] };
+
+  let plists: string[];
+  try {
+    plists = fs
+      .readdirSync(agentsDir)
+      .filter((f) => /^com\.nanoclaw.*\.plist$/.test(f))
+      .map((f) => path.join(agentsDir, f));
+  } catch {
+    return result;
+  }
+
+  const uid = process.getuid?.() ?? 0;
+
+  for (const plistPath of plists) {
+    const label = path.basename(plistPath, '.plist');
+    if (label === ownLabel) continue;
+
+    const status = probeLaunchdPeer(label, plistPath, uid);
+    if (!status) continue;
+    result.checked.push(status);
+
+    if (!status.unhealthy) continue;
+
+    try {
+      execFileSync('launchctl', ['unload', plistPath], { stdio: 'pipe' });
+      log.info('Unloaded unhealthy peer launchd service', {
+        label,
+        state: status.state,
+        runs: status.runs,
+        plistPath,
+      });
+      result.unloaded.push(status);
+    } catch (err) {
+      const message = err instanceof Error ? err.message : String(err);
+      log.warn('Failed to unload peer launchd service', { label, err: message });
+      result.failures.push({ label, err: message });
+    }
+  }
+
+  return result;
+}
+
+function probeLaunchdPeer(label: string, plistPath: string, uid: number): PeerStatus | null {
+  let output: string;
+  try {
+    output = execFileSync('launchctl', ['print', `gui/${uid}/${label}`], {
+      stdio: ['ignore', 'pipe', 'pipe'],
+      encoding: 'utf-8',
+    });
+  } catch {
+    // Not loaded → not currently a threat. Skip silently.
+    return null;
+  }
+
+  const state = /^\s*state\s*=\s*(.+?)\s*$/m.exec(output)?.[1] ?? 'unknown';
+  const runsStr = /^\s*runs\s*=\s*(\d+)/m.exec(output)?.[1];
+  const runs = runsStr ? parseInt(runsStr, 10) : 0;
+
+  const unhealthy = state !== 'running' && runs > UNHEALTHY_RUNS_THRESHOLD;
+  return { label, configPath: plistPath, state, runs, unhealthy };
+}
+
+// ---- systemd (Linux) --------------------------------------------------------
+
+function cleanupSystemdPeers(projectRoot: string): PeerCleanupResult {
+  const ownUnit = getSystemdUnit(projectRoot);
+  const unitDir = path.join(os.homedir(), '.config', 'systemd', 'user');
+  const result: PeerCleanupResult = { checked: [], unloaded: [], failures: [] };
+
+  let units: string[];
+  try {
+    units = fs
+      .readdirSync(unitDir)
+      .filter((f) => /^nanoclaw.*\.service$/.test(f))
+      .map((f) => f.replace(/\.service$/, ''));
+  } catch {
+    return result;
+  }
+
+  for (const unit of units) {
+    if (unit === ownUnit) continue;
+
+    const status = probeSystemdPeer(unit);
+    if (!status) continue;
+    result.checked.push(status);
+
+    if (!status.unhealthy) continue;
+
+    try {
+      execFileSync('systemctl', ['--user', 'disable', '--now', `${unit}.service`], { stdio: 'pipe' });
+      log.info('Disabled unhealthy peer systemd unit', {
+        unit,
+        state: status.state,
+        runs: status.runs,
+      });
+      result.unloaded.push(status);
+    } catch (err) {
+      const message = err instanceof Error ? err.message : String(err);
+      log.warn('Failed to disable peer systemd unit', { unit, err: message });
+      result.failures.push({ label: unit, err: message });
+    }
+  }
+
+  return result;
+}
+
+function probeSystemdPeer(unit: string): PeerStatus | null {
+  const unitPath = path.join(os.homedir(), '.config', 'systemd', 'user', `${unit}.service`);
+  try {
+    const output = execFileSync(
+      'systemctl',
+      ['--user', 'show', '--property=ActiveState,NRestarts', `${unit}.service`],
+      { stdio: ['ignore', 'pipe', 'pipe'], encoding: 'utf-8' },
+    );
+    const activeState = /^ActiveState=(.+)$/m.exec(output)?.[1]?.trim() ?? 'unknown';
+    const restartsStr = /^NRestarts=(\d+)/m.exec(output)?.[1];
+    const runs = restartsStr ? parseInt(restartsStr, 10) : 0;
+
+    const unhealthy =
+      activeState === 'failed' || (activeState !== 'active' && runs > UNHEALTHY_RUNS_THRESHOLD);
+    return { label: unit, configPath: unitPath, state: activeState, runs, unhealthy };
+  } catch {
+    return null;
+  }
+}

setup/service.ts

@@ -11,6 +11,7 @@ import path from 'path';
 
 import { log } from '../src/log.js';
 import { getLaunchdLabel, getSystemdUnit } from '../src/install-slug.js';
+import { cleanupUnhealthyPeers } from './peer-cleanup.js';
 import {
   commandExists,
   getPlatform,
@@ -53,6 +54,19 @@ export async function run(_args: string[]): Promise<void> {
 
   fs.mkdirSync(path.join(projectRoot, 'logs'), { recursive: true });
 
+  // Peer preflight — a crash-looping peer install (most often the legacy v1
+  // `com.nanoclaw` plist) will keep trashing this install's containers on
+  // every respawn via its own cleanupOrphans. Detect and unload any peer
+  // that's unhealthy before we install our service. Healthy peers are left
+  // alone now that container reaping is install-label-scoped.
+  const peerReport = cleanupUnhealthyPeers(projectRoot);
+  if (peerReport.unloaded.length > 0) {
+    log.warn('Unloaded unhealthy peer NanoClaw services', {
+      count: peerReport.unloaded.length,
+      labels: peerReport.unloaded.map((p) => p.label),
+    });
+  }
+
   if (platform === 'macos') {
     setupLaunchd(projectRoot, nodePath, homeDir);
   } else if (platform === 'linux') {

src/config.ts

@@ -2,7 +2,7 @@ import os from 'os';
 import path from 'path';
 
 import { readEnvFile } from './env.js';
-import { getContainerImageBase, getDefaultContainerImage } from './install-slug.js';
+import { getContainerImageBase, getDefaultContainerImage, getInstallSlug } from './install-slug.js';
 import { isValidTimezone } from './timezone.js';
 
 // Read config values from .env (falls back to process.env).
@@ -27,6 +27,10 @@ export const DATA_DIR = path.resolve(PROJECT_ROOT, 'data');
 // `nanoclaw-agent:latest` and clobber each other on rebuild.
 export const CONTAINER_IMAGE_BASE = process.env.CONTAINER_IMAGE_BASE || getContainerImageBase(PROJECT_ROOT);
 export const CONTAINER_IMAGE = process.env.CONTAINER_IMAGE || getDefaultContainerImage(PROJECT_ROOT);
+// Install slug — stamped onto every spawned container via --label so
+// cleanupOrphans only reaps containers from this install, not peers.
+export const INSTALL_SLUG = getInstallSlug(PROJECT_ROOT);
+export const CONTAINER_INSTALL_LABEL = `nanoclaw-install=${INSTALL_SLUG}`;
 export const CONTAINER_TIMEOUT = parseInt(process.env.CONTAINER_TIMEOUT || '1800000', 10);
 export const CONTAINER_MAX_OUTPUT_SIZE = parseInt(process.env.CONTAINER_MAX_OUTPUT_SIZE || '10485760', 10); // 10MB default
 export const ONECLI_URL = process.env.ONECLI_URL || envConfig.ONECLI_URL;

src/container-runner.ts

@@ -12,6 +12,7 @@ import { OneCLI } from '@onecli-sh/sdk';
 import {
   CONTAINER_IMAGE,
   CONTAINER_IMAGE_BASE,
+  CONTAINER_INSTALL_LABEL,
   DATA_DIR,
   GROUPS_DIR,
   ONECLI_API_KEY,
@@ -389,7 +390,7 @@ async function buildContainerArgs(
   providerContribution: ProviderContainerContribution,
   agentIdentifier?: string,
 ): Promise<string[]> {
-  const args: string[] = ['run', '--rm', '--name', containerName];
+  const args: string[] = ['run', '--rm', '--name', containerName, '--label', CONTAINER_INSTALL_LABEL];
 
   // Environment — only vars read by code we don't own.
   // Everything NanoClaw-specific is in container.json (read by runner at startup).

src/container-runtime.test.ts

@@ -24,6 +24,7 @@ import {
   ensureContainerRuntimeRunning,
   cleanupOrphans,
 } from './container-runtime.js';
+import { CONTAINER_INSTALL_LABEL } from './config.js';
 import { log } from './log.js';
 
 beforeEach(() => {
@@ -84,6 +85,17 @@ describe('ensureContainerRuntimeRunning', () => {
 // --- cleanupOrphans ---
 
 describe('cleanupOrphans', () => {
+  it('filters ps by the install label so peers are not reaped', () => {
+    mockExecSync.mockReturnValueOnce('');
+
+    cleanupOrphans();
+
+    expect(mockExecSync).toHaveBeenCalledWith(
+      `${CONTAINER_RUNTIME_BIN} ps --filter label=${CONTAINER_INSTALL_LABEL} --format '{{.Names}}'`,
+      expect.any(Object),
+    );
+  });
+
   it('stops orphaned nanoclaw containers', () => {
     // docker ps returns container names, one per line
     mockExecSync.mockReturnValueOnce('nanoclaw-group1-111\nnanoclaw-group2-222\n');

src/container-runtime.ts

@@ -5,6 +5,7 @@
 import { execSync } from 'child_process';
 import os from 'os';
 
+import { CONTAINER_INSTALL_LABEL } from './config.js';
 import { log } from './log.js';
 
 /** The container runtime binary name. */
@@ -56,13 +57,22 @@ export function ensureContainerRuntimeRunning(): void {
   }
 }
 
-/** Kill orphaned NanoClaw containers from previous runs. */
+/**
+ * Kill orphaned NanoClaw containers from THIS install's previous runs.
+ *
+ * Scoped by label `nanoclaw-install=<slug>` so a crash-looping peer install
+ * cannot reap our containers, and we cannot reap theirs. The label is
+ * stamped onto every container at spawn time — see container-runner.ts.
+ */
 export function cleanupOrphans(): void {
   try {
-    const output = execSync(`${CONTAINER_RUNTIME_BIN} ps --filter name=nanoclaw- --format '{{.Names}}'`, {
-      stdio: ['pipe', 'pipe', 'pipe'],
-      encoding: 'utf-8',
-    });
+    const output = execSync(
+      `${CONTAINER_RUNTIME_BIN} ps --filter label=${CONTAINER_INSTALL_LABEL} --format '{{.Names}}'`,
+      {
+        stdio: ['pipe', 'pipe', 'pipe'],
+        encoding: 'utf-8',
+      },
+    );
     const orphans = output.trim().split('\n').filter(Boolean);
     for (const name of orphans) {
       try {