@crvouga/dev-fleet 0.1.0 → 0.3.0

package/src/worker-host-checks.ts

@@ -4,6 +4,7 @@ import { verifyUbuntuBoot } from './platform/ubuntu-verify';
 import { verifyResourceSafety } from './platform/resource-verify';
 import policy from '../ansible/ubuntu/policy.json';
 import { userInfo } from 'node:os';
+import { T3_HEALTH_SCRIPT, runningVersionCheck } from './platform/t3-service';
 import type { Check, Runner } from './types';
 
 function check(id: string, ok: boolean, summary: string): Check {
@@ -62,6 +63,11 @@ export async function workerHostChecks(
         'Exact declarative Ubuntu package version.'
       )
     ),
+    check(
+      'package:tailscale',
+      audit.packages['tailscale'] === policy.tailscale.version,
+      'Exact declarative Tailscale version.'
+    ),
   ];
   for (const name of policy.alwaysOnServices) {
     const unit = await run([
@@ -115,18 +121,13 @@ async function recoveryChecks(run: Runner): Promise<Check[]> {
     )
   );
   checks.push(await t3HealthCheck(run));
+  checks.push(await runningVersionCheck(run, userInfo().uid));
   const renewal = await run([
     'systemctl',
     '--user',
     'is-active',
     'personal-worker-auth.timer',
   ]);
-  const recovery = await run([
-    'systemctl',
-    '--user',
-    'is-active',
-    'personal-worker-recovery.timer',
-  ]);
   checks.push(
     check(
       'credential-renewal',
@@ -134,17 +135,44 @@ async function recoveryChecks(run: Runner): Promise<Check[]> {
       'Worker credential renewal timer is active.'
     )
   );
-  checks.push(
-    check(
-      'service-recovery',
-      recovery.ok,
-      'Worker service recovery timer is active.'
-    )
-  );
+  checks.push(await serviceRecoveryCheck(run));
   checks.push(await powerRecoveryCheck(run));
   return checks;
 }
 
+/** The recovery timer is active, the script its unit runs exists, and the
+ * last run of the unit succeeded: a watchdog that fails on every tick is not
+ * recovery. */
+export async function serviceRecoveryCheck(run: Runner): Promise<Check> {
+  const timer = await run([
+    'systemctl',
+    '--user',
+    'is-active',
+    'personal-worker-recovery.timer',
+  ]);
+  const script = await run(['test', '-r', T3_HEALTH_SCRIPT]);
+  const unit = await run([
+    'systemctl',
+    '--user',
+    'show',
+    'personal-worker-recovery.service',
+    '--property=Result',
+  ]);
+  const result = /^Result=(.*)$/m.exec(unit.stdout)?.[1] ?? 'unknown';
+  const summary = !timer.ok
+    ? 'WORKER_RECOVERY_TIMER_INACTIVE'
+    : !script.ok
+      ? 'WORKER_RECOVERY_SCRIPT_MISSING'
+      : !unit.ok || result !== 'success'
+        ? `WORKER_RECOVERY_LAST_RUN_FAILED (${/^[\w-]+$/.test(result) ? result : 'unknown'})`
+        : 'Worker service recovery timer is active and its last run succeeded.';
+  return check(
+    'service-recovery',
+    timer.ok && script.ok && unit.ok && result === 'success',
+    summary
+  );
+}
+
 /** The server answers its local API now, not merely that a process exists. */
 async function t3HealthCheck(run: Runner): Promise<Check> {
   const probe = await run(

package/src/workstation-checks.ts

@@ -22,7 +22,7 @@ export async function workstationChecks(
           summary: 'The devfleet launcher is missing.',
           remedy: {
             action:
-              'Run fleet-sync from the manager, or sh bootstrap/install-launcher.sh in the dev-fleet checkout on this computer.',
+              'Run fleet-sync from the manager: it links the launcher to the checkout of the revision it applies.',
           },
         },
   ];

package/src/index.ts

@@ -1,6 +0,0 @@
-export { parseConfig, loadConfig, selectNodes } from './config';
-export { doctor } from './manager';
-export { probe } from './probe';
-export { makeReport, reportExitCode, renderReport } from './report';
-export { syncPlan, syncWorkspace } from './sync';
-export type * from './types';

package/src/test-fixtures.ts

@@ -1,36 +0,0 @@
-import { mkdir, mkdtemp, writeFile } from 'node:fs/promises';
-import { tmpdir } from 'node:os';
-import { join } from 'node:path';
-import type { CommandResult, Runner, FleetNode } from './types';
-
-export const HEAD = 'a'.repeat(40);
-export const NODE: FleetNode = {
-  id: 'dev-one',
-  ssh: 'dev-one',
-  workspace: '/home/dev/workspace',
-  required: [],
-};
-export const success = (stdout = ''): CommandResult => ({
-  ok: true,
-  stdout,
-  stderr: '',
-  reason: 'exit',
-});
-export const failure = (stderr = ''): CommandResult => ({
-  ok: false,
-  stdout: '',
-  stderr,
-  reason: 'exit',
-});
-export const gitRunner: Runner = async (command) =>
-  success(command.includes('rev-parse') ? HEAD : '');
-
-export async function workspace(): Promise<string> {
-  const root = await mkdtemp(join(tmpdir(), "fleet test's "));
-  for (const file of ['package.json', 'bun.lock', 'mise.toml', 'mise.lock'])
-    await writeFile(join(root, file), '{}');
-  await mkdir(join(root, 'packages/example'), { recursive: true });
-  await writeFile(join(root, 'packages/example/package.json'), '{}');
-  await mkdir(join(root, '.mise/locks'), { recursive: true });
-  return root;
-}