@yadurajfleetos/cli 0.3.0 → 0.4.1

package/README.md

@@ -16,7 +16,7 @@ npx @yadurajfleetos/cli --help
 
 ### Build and Install from Source
 ```bash
-git clone https://github.com/YadurajManu/BhaiMerra.git fleet-os
+git clone https://github.com/YadurajManu/fleet-os.git fleet-os
 cd fleet-os/cli
 npm install
 npm run build

package/dist/api.js

@@ -89,13 +89,32 @@ export async function requireFleet(explicit) {
     if (explicit)
         return explicit;
     const profile = await loadProfile();
-    if (profile.fleetId)
-        return profile.fleetId;
+    // A cached id is a hint, not a fact. It survives the fleet being deleted, the
+    // control plane being rebuilt, and signing in as somebody else — and the
+    // failure was a bare "Fleet not found" on every command, which names neither
+    // the cache nor the way out. Verify it, and fall through when it is stale.
+    if (profile.fleetId) {
+        try {
+            await request('GET', `/fleets/${profile.fleetId}`);
+            return profile.fleetId;
+        }
+        catch (err) {
+            if (!(err instanceof CliError) || err.exitCode !== EXIT.failure)
+                throw err;
+            await saveProfile({ ...profile, fleetId: undefined, fleetName: undefined });
+        }
+    }
     const { body } = await request('GET', '/fleets');
-    if (body.fleets.length === 1)
+    if (body.fleets.length === 1) {
+        // Remember it, so the next command does not pay for this lookup again.
+        await saveProfile({ ...profile, fleetId: body.fleets[0].id, fleetName: body.fleets[0].name });
         return body.fleets[0].id;
-    if (!body.fleets.length)
-        throw new CliError('You have no fleets yet.', EXIT.usage);
+    }
+    if (!body.fleets.length) {
+        throw new CliError('This account owns no fleets on ' + (profile.api ?? 'this control plane') + '.\n' +
+            '  If you expected one, you may be signed in as the wrong account — check with `fleet auth whoami`,\n' +
+            '  or create a fleet in the dashboard and run `fleet use <name>`.', EXIT.usage);
+    }
     throw new CliError(`You are in several fleets. Pass --fleet <id> or run \`fleet use <name>\`:\n` +
         body.fleets.map((f) => `  ${f.name}  ${f.id}`).join('\n'), EXIT.usage);
 }

package/dist/commands/auth.js

@@ -232,8 +232,58 @@ export const authCommand = {
                 ]));
                 return;
             }
+            /**
+             * Ask for a reset link. The control plane answers 204 for every address,
+             * known or not, so this command cannot be used to discover whether an
+             * account exists either - and says so, rather than implying the mail is
+             * definitely on its way.
+             */
+            case 'forgot': {
+                if (!profile.api) {
+                    profile.api = validApi(await requiredPrompt('control plane URL', { hint: 'e.g. https://fleetapi.example.com' }));
+                }
+                const email = typeof flags.email === 'string' ? flags.email : await requiredPrompt('email');
+                await task('requesting a reset link', async () => {
+                    // auth:false - you cannot be signed in when you have forgotten the password.
+                    await request('POST', '/auth/forgot', { profile, body: { email }, auth: false });
+                });
+                console.log(`\n${glyph.ok} If an account exists for ${c.bold(email)}, a reset link is on its way.`);
+                console.log(c.dim('  The link works once and expires in 30 minutes.'));
+                console.log(c.dim(`\n  fleet auth reset --token <token>   finish it here`));
+                console.log(c.dim('  or open the link in a browser instead'));
+                return;
+            }
+            /**
+             * Finish a reset without a browser. The token comes from the email; the
+             * new password is prompted for rather than passed as a flag, because a
+             * flag lands in shell history.
+             */
+            case 'reset': {
+                if (!profile.api) {
+                    profile.api = validApi(await requiredPrompt('control plane URL'));
+                }
+                const token = typeof flags.token === 'string'
+                    ? flags.token
+                    : await requiredPrompt('reset token', {
+                        hint: 'the token from the reset email, or the token= part of the link',
+                    });
+                const password = await requiredPrompt('new password', { silent: true });
+                if (password.length < 12) {
+                    throw new CliError('Password must be at least 12 characters.', EXIT.usage);
+                }
+                const again = await requiredPrompt('confirm password', { silent: true });
+                if (password !== again)
+                    throw new CliError('Those did not match.', EXIT.usage);
+                await task('setting your new password', async () => {
+                    await request('POST', '/auth/reset', { profile, body: { token, password }, auth: false });
+                });
+                console.log(`\n${glyph.ok} ${c.green('password changed')}`);
+                console.log(c.dim('  Every other session was signed out.'));
+                console.log(c.dim(`\n  fleet auth login   sign in again`));
+                return;
+            }
             default:
-                throw new CliError('usage: fleet auth login|logout|whoami', EXIT.usage);
+                throw new CliError('usage: fleet auth login|logout|whoami|forgot|reset', EXIT.usage);
         }
     },
 };

package/dist/commands/backups.js

@@ -0,0 +1,122 @@
+/**
+ * fleet backup — take a copy of a service's volume, and list what exists.
+ *
+ * A volume is the one thing Fleet cannot reproduce. Everything else here is
+ * derived: an image rebuilds from a commit, a container recreates from a
+ * manifest. The bytes in a database's data directory live on exactly one disk,
+ * and until now there was no way to get a copy of them off it.
+ */
+import { request, requireFleet, CliError, EXIT } from '../api.js';
+import { c, table, relativeTime } from '../render.js';
+import { glyph, task } from '../ui.js';
+export function humanBytes(n) {
+    if (n < 1024)
+        return `${n} B`;
+    const units = ['KB', 'MB', 'GB', 'TB'];
+    let value = n / 1024;
+    let i = 0;
+    while (value >= 1024 && i < units.length - 1) {
+        value /= 1024;
+        i++;
+    }
+    return `${value.toFixed(value >= 10 ? 0 : 1)} ${units[i]}`;
+}
+async function resolveService(fleetId, name) {
+    const { body } = await request('GET', `/fleets/${fleetId}/services`);
+    const match = body.services.find((s) => s.name === name || s.id === name);
+    if (!match) {
+        throw new CliError(`No service called "${name}". Known: ${body.services.map((s) => s.name).join(', ') || 'none'}`, EXIT.usage);
+    }
+    return match;
+}
+const STATUS_TONE = {
+    complete: c.green,
+    running: c.yellow,
+    pending: c.dim,
+    failed: c.red,
+};
+export const backupCommand = {
+    async run(args, flags) {
+        const [name] = args;
+        if (!name)
+            throw new CliError('usage: fleet backup <service>', EXIT.usage);
+        const fleetId = await requireFleet(typeof flags.fleet === 'string' ? flags.fleet : undefined);
+        const service = await resolveService(fleetId, name);
+        const created = await task(`asking ${c.bold(service.name)}'s node for a copy of its volume`, async () => (await request('POST', `/fleets/${fleetId}/services/${service.id}/backups`, { body: {} })).body.backup, { done: (b) => `queued backup of ${b.volumeRef}` });
+        console.log(`\n${glyph.ok} ${c.green('queued')}  ${c.bold(created.id.slice(0, 8))}`);
+        console.log(c.dim('  The node holding the volume performs it on its next poll — a large volume takes a while.'));
+        console.log(c.dim(`\n  fleet backups ${service.name}   watch it finish`));
+    },
+};
+export const backupsCommand = {
+    async run(args, flags) {
+        const [name] = args;
+        if (!name)
+            throw new CliError('usage: fleet backups <service>', EXIT.usage);
+        const fleetId = await requireFleet(typeof flags.fleet === 'string' ? flags.fleet : undefined);
+        const service = await resolveService(fleetId, name);
+        const { body } = await request('GET', `/fleets/${fleetId}/services/${service.id}/backups`);
+        if (flags.json)
+            return console.log(JSON.stringify(body.backups, null, 2));
+        if (!body.backups.length) {
+            if (!service.persistentVolume) {
+                console.log(`"${service.name}" has no volume, so there is nothing to back up.`);
+                console.log(c.dim('  Its image and manifest already describe everything it holds.'));
+                return;
+            }
+            console.log(`No backups of ${c.bold(service.name)} yet.`);
+            console.log(c.dim(`  take one with \`fleet backup ${service.name}\``));
+            return;
+        }
+        console.log(table(['when', 'status', 'size', 'source', 'id'], body.backups.map((b) => [
+            relativeTime(b.createdAt),
+            STATUS_TONE[b.status](b.status),
+            b.sizeBytes ? humanBytes(b.sizeBytes) : c.dim('—'),
+            b.scheduled ? c.dim('scheduled') : c.dim('manual'),
+            c.dim(b.id.slice(0, 8)),
+        ])));
+        // Failures are the half people come here for, and a table cell is too
+        // narrow to say anything useful about one.
+        const failed = body.backups.filter((b) => b.status === 'failed' && b.failureReason);
+        for (const b of failed.slice(0, 3)) {
+            console.log(`\n${glyph.warn} ${c.yellow(b.id.slice(0, 8))}  ${b.failureReason.split('\n')[0].slice(0, 160)}`);
+        }
+        const complete = body.backups.filter((b) => b.status === 'complete');
+        if (complete.length) {
+            const total = complete.reduce((sum, b) => sum + (b.sizeBytes ?? 0), 0);
+            console.log(c.dim(`\n  ${complete.length} complete, ${humanBytes(total)} stored.`));
+        }
+    },
+};
+export const restoreCommand = {
+    async run(args, flags) {
+        const [name, which] = args;
+        if (!name)
+            throw new CliError('usage: fleet restore <service> [backup-id]', EXIT.usage);
+        const fleetId = await requireFleet(typeof flags.fleet === 'string' ? flags.fleet : undefined);
+        const service = await resolveService(fleetId, name);
+        const { body } = await request('GET', `/fleets/${fleetId}/services/${service.id}/backups`);
+        const complete = body.backups.filter((b) => b.status === 'complete');
+        if (!complete.length) {
+            throw new CliError(`No completed backup of "${service.name}" to restore. \`fleet backups ${service.name}\` shows what exists.`, EXIT.usage);
+        }
+        // Named by prefix, or the most recent — which is what "restore it" almost
+        // always means, and typing a uuid from a table is nobody's idea of a good
+        // recovery experience.
+        const target = which
+            ? complete.find((b) => b.id === which || b.id.startsWith(which))
+            : complete[0];
+        if (!target) {
+            throw new CliError(`No completed backup of "${service.name}" starting with "${which}".`, EXIT.usage);
+        }
+        console.log(`\n  Restoring ${c.bold(service.name)} from ${c.bold(target.id.slice(0, 8))}` +
+            ` ${c.dim(`(${relativeTime(target.createdAt)}, ${target.sizeBytes ? humanBytes(target.sizeBytes) : 'unknown size'})`)}`);
+        console.log(c.dim('  The archive is written into the volume, over whatever is there now.\n' +
+            '  The service must be stopped: writing a data directory underneath a\n' +
+            '  running process corrupts it, so this is refused while it serves.\n'));
+        const started = await task('queueing the restore', async () => (await request('POST', `/fleets/${fleetId}/backups/${target.id}/restore`, { body: {} })).body.restore, { done: () => 'queued' });
+        console.log(`\n${glyph.ok} ${c.green('queued')}  ${c.bold(started.id.slice(0, 8))}`);
+        console.log(c.dim('  The node writes it into the volume on its next poll.'));
+        console.log(c.dim(`\n  fleet deploy ${service.name}   bring it back up once the restore lands`));
+    },
+};

package/dist/commands/doctor.js

@@ -3,12 +3,48 @@ import { loadProfile } from '../config.js';
 import { c, relativeTime } from '../render.js';
 import { glyph, rule, task } from '../ui.js';
 const icon = (state) => state === 'ok' ? glyph.ok : state === 'warn' ? glyph.warn : glyph.fail;
+/**
+ * A build failure carries the whole buildx transcript. One summary line
+ * belongs in a health report; `fleet deployments` is where the rest lives.
+ */
+function firstLine(text) {
+    const line = text.split('\n').find((l) => l.trim()) ?? text;
+    return line.length > 160 ? `${line.slice(0, 157)}…` : line;
+}
+/**
+ * Statuses an edge returns when it could not reach the origin at all.
+ * Cloudflare's 52x range and 530 mean the tunnel is down; 502/503/504 mean the
+ * same thing from any reverse proxy.
+ */
+const ORIGIN_UNREACHABLE = new Set([502, 503, 504, 520, 521, 522, 523, 524, 525, 526, 527, 530]);
+/**
+ * Is the service reachable through the ingress?
+ *
+ * That question is not "did it return 200". An API with no route at `/` answers
+ * 404, and a 404 from the *application* is proof the whole chain works —
+ * ingress, tunnel, agent, container. Reporting that as a failure sends people
+ * to check DNS and node health when nothing is wrong, which is exactly what it
+ * did for an API whose /health returned 200 the whole time.
+ *
+ * What does indicate a broken path is the edge answering on the origin's
+ * behalf, or nothing answering at all.
+ */
 async function reach(url) {
     try {
         const response = await fetch(url, { redirect: 'manual', signal: AbortSignal.timeout(8_000) });
-        return response.status >= 200 && response.status < 400
-            ? { ok: true, detail: `HTTPS answered ${response.status}` }
-            : { ok: false, detail: `HTTPS answered ${response.status}` };
+        const status = response.status;
+        if (ORIGIN_UNREACHABLE.has(status)) {
+            return { ok: false, detail: `HTTPS answered ${status} — the edge could not reach the container` };
+        }
+        if (status >= 200 && status < 400) {
+            return { ok: true, detail: `HTTPS answered ${status}` };
+        }
+        if (status < 500) {
+            // The application answered. It has an opinion about the request, which
+            // means everything in front of it is working.
+            return { ok: true, detail: `HTTPS answered ${status} — reachable; the app has no route there` };
+        }
+        return { ok: false, detail: `HTTPS answered ${status} — reachable, but the app is erroring` };
     }
     catch (error) {
         return { ok: false, detail: error instanceof Error ? error.message : String(error) };
@@ -93,12 +129,24 @@ export const doctorCommand = {
                     checks.push({ state: 'fail', label: `reconcile ${node.name}`, detail: runtime.lastReconcileError, remedy: 'Run `fleet logs <service> --follow` and inspect the deployment history.' });
             }
         }
-        const failed = result.deploymentHistory.flatMap(({ service, deployments }) => deployments.filter((deployment) => deployment.status === 'failed' || deployment.failureReason).slice(0, 1).map((deployment) => ({ service: service.name, deployment })));
+        // Only the *current* deployment of each service can be a current problem.
+        // Scanning the whole history meant a service that failed once and then
+        // deployed successfully still reported the old failure, so a healthy fleet
+        // showed a wall of buildx output from a build that had since been redone.
+        const failed = result.deploymentHistory.flatMap(({ service, deployments }) => {
+            const latest = deployments[0];
+            if (!latest)
+                return [];
+            const isFailure = latest.status === 'failed' || Boolean(latest.failureReason);
+            return isFailure ? [{ service: service.name, deployment: latest }] : [];
+        });
         checks.push(failed.length
             ? {
                 state: 'fail',
                 label: 'deployments',
-                detail: failed.map(({ service, deployment }) => `${service}: ${deployment.failureReason ?? deployment.status}`).join('; '),
+                detail: failed
+                    .map(({ service, deployment }) => `${service}: ${firstLine(deployment.failureReason ?? deployment.status)}`)
+                    .join('; '),
                 remedy: 'Run `fleet deployments <service>` for history and `fleet logs <service> --follow` for the current container tail.',
             }
             : { state: 'ok', label: 'deployments', detail: result.services.length ? 'No recorded deployment failures.' : 'No services declared yet.' });

package/dist/commands/index.js

@@ -9,6 +9,7 @@ import { openCommand } from './open.js';
 import { downCommand } from './down.js';
 import { unpairCommand, agentCommand } from './unpair.js';
 import { secretsCommand } from './secrets.js';
+import { backupCommand, backupsCommand, restoreCommand } from './backups.js';
 import { applyCommand, deployCommand, deploymentsCommand, initCommand, logsCommand, removeServiceCommand, restartCommand, rollbackCommand, rescheduleCommand, servicesCommand, validateCommand, whereCommand, } from './services.js';
 export const commands = {
     up: upCommand,
@@ -35,6 +36,9 @@ export const commands = {
     events: eventsCommand,
     alerts: alertsCommand,
     secrets: secretsCommand,
+    backup: backupCommand,
+    backups: backupsCommand,
+    restore: restoreCommand,
     unpair: unpairCommand,
     agent: agentCommand,
 };

package/dist/index.js

@@ -52,12 +52,16 @@ const GROUPS = [
             ['secrets set <KEY>', 'Store a credential; the value is never echoed or logged'],
             ['secrets import [.env]', 'Store the secrets fleet.yaml declares, read from a .env file'],
             ['secrets rm <KEY>', 'Remove a stored credential'],
+            ['backup <service>', "Copy a service's volume off the node holding it"],
+            ['backups <service>', 'List backups, newest first'],
+            ['restore <service> [id]', 'Write a backup back into the volume; service must be stopped'],
             ['nodes cordon <name>', 'Stop scheduling new work onto a node'],
             ['nodes uncordon <name>', 'Allow scheduling again'],
             ['nodes rm <name>', 'Revoke and remove a node'],
             ['unpair', 'Remove this machine from its fleet, run on the machine'],
             ['alerts', 'List, add, and test alert rules'],
             ['auth login|logout|whoami', 'Sign in to a control plane'],
+            ['auth forgot|reset', 'Recover an account you are locked out of'],
         ],
     ],
 ];

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@yadurajfleetos/cli",
-  "version": "0.3.0",
+  "version": "0.4.1",
   "description": "Fleet OS command-line interface for deploying and orchestrating services on user-owned hardware",
   "type": "module",
   "license": "MIT",