feat(lnd): rotate Lightning macaroons from the dashboard, and stop stranding BTCPay
Demo images / Build & push demo images (push) Successful in 3m34s

Rotating LND's macaroons was an SSH-only script, which in practice meant it did
not happen — while a macaroon is a bearer token with no revocation and no expiry,
so anything that ever read one keeps the ability to spend until they are
replaced. Settings → Lightning credentials now does it behind the node password,
shows a step checklist, and refuses to report success unless it has confirmed the
node identity and channel census are unchanged.

Three findings from performing a real rotation on a dev node, each fixed here:

1. BTCPay was left holding a dead credential, silently. Its connection string
   carries the macaroon INLINE (LND's datadir is owned by its container subuid,
   so btcpay cannot bind-mount the file), and the daemon only regenerates that
   secret when LND's TLS cert thumbprint changes — which macaroon rotation does
   not touch. Result: btcpay up, LND up, both healthy, every Lightning payment
   failing, nothing anywhere saying why.

2. Rewriting the secret is not enough to fix it. `secret_env_hash` makes the
   change visible as env drift, but the reconcile loop runs `ExistingOnly` at
   boot AND periodically, and there it deliberately leaves running
   restart-sensitive apps untouched — observed once per tick for half an hour on
   the dev node. So this reuses FED-07's `credential_rotated` carve-out via a new
   default-no-op `ContainerOrchestrator::mark_credential_rotated`, on the same
   reasoning: restart sensitivity protects apps that are working, and this one is
   working only in appearance. The shell script cannot reach an in-process flag,
   so it removes the container and lets desired-state recovery rebuild it.

3. LND stayed locked forever on a loaded node. The unlocker is only served after
   channel.db/graph.db/wallet.db open, measured at 2m38s on a box running 30
   containers; the unlock helper gave up at ~60s. That is not a harmless retry —
   reconcile records the post-start hook as failed, restarts LND, and the slow
   open begins again, so the wallet never opens and every LND-dependent app stays
   broken. The not-ready budget is now ~10 minutes; a genuinely wrong password
   still exits on the first pass via `all_rejected`.

Safety properties worth not regressing:
- No macaroon content in any response, error, log line or the polled progress
  feed — digests and byte counts only.
- Rotation unlocks via a new `unlock_existing_wallet_no_wipe`, so there is no
  code path from "rotate my credentials" to `recreate_wallet_destructively`. A
  wallet whose password this node lacks fails the rotation with the wallet intact.
- Channels are compared as active+inactive totals, not `num_active_channels`,
  which legitimately dips after any restart while peers reconnect.
- Backup verified by file count before anything is deleted.

Verified: cargo check + fmt clean, 6 new unit tests and the 6 existing
container::lnd tests pass, vue-tsc clean, and the built bundle contains the three
new RPC method names (the frontend build can silently no-op).

Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
This commit is contained in:
archipelago
2026-08-08 07:45:51 -04:00
co-authored by Claude Opus 5
parent b7e57ca9cf
commit d15cd58d7f
13 changed files with 1628 additions and 13 deletions
+63
View File
@@ -1158,6 +1158,69 @@ class RPCClient {
})
}
/** This node's Lightning credential state. Digests and counts only — the
* backend never returns macaroon content, so nothing here is sensitive. */
async lndMacaroonStatus(): Promise<LndMacaroonStatus> {
return this.call({ method: 'lnd.macaroon-status', timeout: 30000 })
}
/** Begin a macaroon rotation. Returns as soon as the job is accepted; the
* work takes minutes (LND has to close and reopen its databases), so poll
* `lndMacaroonRotationProgress` for the outcome. */
async lndRotateMacaroons(password: string): Promise<{ status: string }> {
return this.call({
method: 'lnd.rotate-macaroons',
params: { password },
timeout: 30000,
})
}
async lndMacaroonRotationProgress(): Promise<LndRotationProgress> {
return this.call({ method: 'lnd.macaroon-rotation-progress' })
}
}
export type RotationStepState = 'pending' | 'running' | 'done' | 'failed' | 'skipped'
export interface LndRotationStep {
key: string
label: string
state: RotationStepState
detail: string | null
}
export interface LndRotationProgress {
running: boolean
/** null while running, then the verdict. Lets the UI tell "in progress"
* apart from "finished and failed". */
ok: boolean | null
started_at: string | null
finished_at: string | null
error: string | null
steps: LndRotationStep[]
/** Holds the OLD root key, so it is still secret. The UI tells the operator
* to delete it once every wallet app has been re-paired. */
backup_path: string | null
identity_pubkey: string | null
channels_before: number | null
channels_after: number | null
new_admin_macaroon_sha256: string | null
}
export interface LndMacaroonStatus {
installed: boolean
admin_macaroon_sha256: string | null
/** When LND last minted these credentials, host local time. */
issued_at: string | null
identity_pubkey: string | null
channels_open: number | null
channels_pending: number | null
/** Why LND could not be asked, when it could not. */
lnd_error: string | null
btcpay_uses_internal_lnd: boolean
/** null when BTCPay has no internal Lightning node — an absence, not a fault. */
btcpay_credential_current: boolean | null
rotation: LndRotationProgress
}
export const rpcClient = new RPCClient()
@@ -0,0 +1,350 @@
<script setup lang="ts">
import { computed, onMounted, onUnmounted, ref } from 'vue'
import { rpcClient, type LndMacaroonStatus, type LndRotationProgress } from '@/api/rpc-client'
// A Lightning macaroon is a bearer token: whoever holds one can spend from this
// node's wallet. Rotating them is the only way to take that ability back from
// anything that has seen one — a lost phone, a shared screenshot, an app that
// ran a version with a published vulnerability. Until now that meant SSHing in
// and running a script, which in practice meant it did not happen.
//
// The screen is deliberately fact-first. Before anyone clicks the button they
// can see when the credentials were issued, which node they belong to, how many
// channels must survive, and whether anything on this node is already out of
// step — because "will this close my channels?" is the question that stops
// people rotating, and the honest answer is on the page.
const status = ref<LndMacaroonStatus | null>(null)
const loading = ref(true)
const loadError = ref('')
const showConfirm = ref(false)
const password = ref('')
const submitting = ref(false)
const confirmError = ref('')
let poll: ReturnType<typeof setInterval> | null = null
const rotation = computed<LndRotationProgress | null>(() => status.value?.rotation ?? null)
const isRunning = computed(() => rotation.value?.running === true)
/** A finished rotation, successful or not. `ok` is null while running. */
const finished = computed(
() => rotation.value !== null && !rotation.value.running && rotation.value.ok !== null,
)
/** BTCPay embeds a copy of the macaroon inline and cannot self-heal, so it is
* the one dependency that can silently fall out of step. `false` is the state
* worth shouting about; `null` just means BTCPay has no internal node. */
const btcpayStale = computed(() => status.value?.btcpay_credential_current === false)
async function load() {
try {
status.value = await rpcClient.lndMacaroonStatus()
loadError.value = ''
// Poll only while there is something to watch, so an idle Settings tab
// isn't waking the node every few seconds.
if (status.value.rotation.running) startPolling()
else stopPolling()
} catch (e) {
loadError.value = e instanceof Error ? e.message : String(e)
} finally {
loading.value = false
}
}
function startPolling() {
if (poll) return
poll = setInterval(load, 4000)
}
function stopPolling() {
if (poll) {
clearInterval(poll)
poll = null
}
}
function openConfirm() {
password.value = ''
confirmError.value = ''
showConfirm.value = true
}
function closeConfirm() {
showConfirm.value = false
password.value = ''
confirmError.value = ''
}
async function rotate() {
submitting.value = true
confirmError.value = ''
try {
await rpcClient.lndRotateMacaroons(password.value)
closeConfirm()
startPolling()
await load()
} catch (e) {
confirmError.value = e instanceof Error ? e.message : String(e)
} finally {
submitting.value = false
password.value = ''
}
}
function stepIcon(state: string): string {
switch (state) {
case 'done':
return '✓'
case 'failed':
return '✕'
case 'skipped':
return ''
case 'running':
return '…'
default:
return '·'
}
}
function stepClass(state: string): string {
switch (state) {
case 'done':
return 'text-emerald-400'
case 'failed':
return 'text-red-400'
case 'skipped':
return 'text-white/40'
case 'running':
return 'text-orange-300'
default:
return 'text-white/30'
}
}
/** First 16 characters is plenty to compare two digests by eye, and keeps the
* line readable on a phone. */
function shortHash(h: string | null): string {
return h ? `${h.slice(0, 16)}` : '—'
}
onMounted(load)
onUnmounted(stopPolling)
</script>
<template>
<div class="mb-6">
<h3 class="text-base font-medium text-white/90 mb-1">Lightning credentials</h3>
<p class="text-sm text-white/60 mb-4">
Wallet apps like Zeus connect to this node using a Lightning credential a
token that lets them spend. Rotating replaces every one of them, so anything
that copied an old token can no longer use it. Your coins and channels are
not touched: the node keeps its identity and no channel is closed.
</p>
<div v-if="loading" class="text-sm text-white/50">Checking</div>
<div
v-else-if="loadError"
class="p-3 bg-white/5 border border-white/10 rounded-lg text-sm text-white/70"
>
Could not read the Lightning credential state: {{ loadError }}
</div>
<div
v-else-if="!status?.installed"
class="p-3 bg-white/5 border border-white/10 rounded-lg text-sm text-white/70"
>
Lightning is not set up on this node yet, so there are no credentials to
rotate. Install the Lightning app first.
</div>
<div v-else class="space-y-4">
<!-- What exists right now -->
<dl class="grid grid-cols-1 sm:grid-cols-2 gap-3 text-sm">
<div>
<dt class="text-white/50 text-xs">Issued</dt>
<dd class="text-white/80">{{ status.issued_at || 'unknown' }}</dd>
</div>
<div>
<dt class="text-white/50 text-xs">Credential fingerprint</dt>
<dd class="text-white/80 font-mono text-xs break-all">
{{ shortHash(status.admin_macaroon_sha256) }}
</dd>
</div>
<div>
<dt class="text-white/50 text-xs">Channels that must survive</dt>
<dd class="text-white/80">
<template v-if="status.channels_open !== null">
{{ status.channels_open }} open<span v-if="status.channels_pending">
, {{ status.channels_pending }} pending</span
>
</template>
<span v-else class="text-white/50">not readable Lightning is not answering</span>
</dd>
</div>
<div>
<dt class="text-white/50 text-xs">Node identity</dt>
<dd class="text-white/80 font-mono text-xs break-all">
{{ status.identity_pubkey ? `${status.identity_pubkey.slice(0, 16)}` : '—' }}
</dd>
</div>
</dl>
<!-- Lightning has to be answering for a rotation to be verifiable at all,
so this is a blocker rather than a footnote. -->
<div
v-if="status.lnd_error && !isRunning"
class="p-3 bg-orange-500/10 border border-orange-500/30 rounded-lg text-sm text-orange-100/90"
>
<p class="font-medium mb-1">Lightning is not answering right now.</p>
<p class="text-orange-100/70">
Rotation is blocked until it is: without a reading from before the
change there is no way to prove afterwards that your channels came
back. Wait for Lightning to finish starting and reload this page.
</p>
<p class="text-xs text-orange-100/50 mt-2 font-mono break-all">{{ status.lnd_error }}</p>
</div>
<!-- The failure this whole feature exists to prevent. -->
<div
v-if="btcpayStale"
class="p-3 bg-red-500/10 border border-red-500/30 rounded-lg text-sm text-red-100/90"
>
<p class="font-medium mb-1">BTCPay Server is holding an old Lightning credential.</p>
<p class="text-red-100/70">
BTCPay keeps its own copy of the credential, and the copy it has no
longer works so its Lightning payments will fail even though both
apps look healthy. Rotating now repairs this as part of the run.
</p>
</div>
<!-- Progress. Shown while running and kept afterwards, because the
verdict ("same node, same channels") is the reassurance the operator
came here for. -->
<div v-if="rotation && (isRunning || finished)" class="p-3 bg-white/5 border border-white/10 rounded-lg">
<p class="text-sm font-medium text-white/80 mb-2">
<span v-if="isRunning">Rotating</span>
<span v-else-if="rotation.ok" class="text-emerald-400">Rotation complete</span>
<span v-else class="text-red-400">Rotation failed</span>
</p>
<ul class="space-y-1.5">
<li v-for="step in rotation.steps" :key="step.key" class="text-sm">
<span class="font-mono mr-2" :class="stepClass(step.state)">{{
stepIcon(step.state)
}}</span>
<span :class="step.state === 'pending' ? 'text-white/40' : 'text-white/80'">{{
step.label
}}</span>
<p v-if="step.detail" class="ml-6 text-xs text-white/50">{{ step.detail }}</p>
</li>
</ul>
<p v-if="rotation.error" class="mt-3 text-xs text-red-300/90 break-words">
{{ rotation.error }}
</p>
<div v-if="finished && rotation.ok" class="mt-3 space-y-2 text-xs text-white/60">
<p class="text-white/80">
Re-pair anything that connects to this node Zeus most importantly.
Open the Lightning app and scan its pairing QR again; it serves the
new credential.
</p>
<p v-if="rotation.backup_path">
The old credentials were backed up on the node so a mistake is
recoverable. That backup is still sensitive. Once every app is
re-paired, delete it:
<code class="block mt-1 px-2 py-1 bg-black/30 rounded font-mono break-all"
>sudo rm -rf {{ rotation.backup_path }}</code
>
</p>
</div>
</div>
<button
:disabled="isRunning || !!status.lnd_error"
class="w-full flex items-center justify-center gap-2 px-4 py-2 rounded-lg glass-button glass-button-warning font-medium disabled:opacity-50 disabled:cursor-not-allowed"
@click="openConfirm"
>
<svg class="w-5 h-5" fill="none" stroke="currentColor" viewBox="0 0 24 24">
<path
stroke-linecap="round"
stroke-linejoin="round"
stroke-width="2"
d="M4 4v5h.582m15.356 2A8.001 8.001 0 004.582 9m0 0H9m11 11v-5h-.581m0 0a8.003 8.003 0 01-15.357-2m15.357 2H15"
/>
</svg>
<span>{{ isRunning ? 'Rotating…' : 'Rotate Lightning credentials' }}</span>
</button>
</div>
</div>
<!-- Confirmation. Teleported to body: a glass-panel ancestor creates a
transform context that would trap a position:fixed backdrop. -->
<Teleport to="body">
<div
v-if="showConfirm"
class="fixed inset-0 z-50 flex items-center justify-center p-4 bg-black/60 backdrop-blur-md"
@click.self="closeConfirm"
@keydown.escape="closeConfirm"
>
<div
class="glass-card p-6 max-w-md w-full"
role="dialog"
aria-modal="true"
aria-labelledby="rotate-macaroon-title"
>
<h3 id="rotate-macaroon-title" class="text-lg font-semibold text-white mb-2">
Rotate Lightning credentials
</h3>
<div class="text-sm text-white/70 space-y-2 mb-4">
<p>
<strong class="text-white/90">What changes:</strong> every app paired
with this node stops working until you re-pair it. Zeus and any other
remote wallet will need to scan a fresh pairing code.
</p>
<p>
<strong class="text-white/90">What does not:</strong> your coins and
your channels. The node keeps its identity, nothing is closed, and
this run refuses to report success unless it has confirmed both.
</p>
<p>
Lightning restarts as part of this, which takes a few minutes on a
busy node. Payments cannot be sent or received during that window.
</p>
</div>
<form class="space-y-4" @submit.prevent="rotate">
<label class="block">
<span class="text-xs text-white/60">Confirm with your node password</span>
<input
v-model="password"
type="password"
required
autocomplete="current-password"
class="mt-1 w-full px-3 py-2 rounded-lg bg-white/10 text-white border border-white/20 focus:border-orange-500 focus:ring-1 focus:ring-orange-500"
placeholder="Node password"
/>
</label>
<p v-if="confirmError" class="text-sm text-red-400 break-words">{{ confirmError }}</p>
<div class="flex gap-3">
<button
type="button"
class="flex-1 px-4 py-2 rounded-lg glass-button font-medium"
@click="closeConfirm"
>
Cancel
</button>
<button
type="submit"
:disabled="submitting || !password"
class="flex-1 px-4 py-2 rounded-lg bg-orange-500 text-white font-medium hover:bg-orange-600 disabled:opacity-50 disabled:cursor-not-allowed transition-colors"
>
{{ submitting ? 'Starting…' : 'Rotate' }}
</button>
</div>
</form>
</div>
</div>
</Teleport>
</template>
@@ -6,6 +6,7 @@ import AIDataAccessSection from '@/views/settings/AIDataAccessSection.vue'
import WebhookSection from '@/views/settings/WebhookSection.vue'
import TelemetrySection from '@/views/settings/TelemetrySection.vue'
import NodeCertificateSection from '@/views/settings/NodeCertificateSection.vue'
import LightningCredentialsSection from '@/views/settings/LightningCredentialsSection.vue'
import BackupSection from '@/views/settings/BackupSection.vue'
import SystemDangerZone from '@/views/settings/SystemDangerZone.vue'
</script>
@@ -18,6 +19,7 @@ import SystemDangerZone from '@/views/settings/SystemDangerZone.vue'
<WebhookSection />
<TelemetrySection />
<NodeCertificateSection />
<LightningCredentialsSection />
<BackupSection />
<SystemDangerZone />
</template>