fix(usage): let an operator clear a participation the collector never accepted

Probing the live collector to settle the delete-sequence question turned up
something else: usage.picpeak.app answers a valid usage.v2 registration with
INVALID_PACKET while the identical v1 flow is accepted. It does not speak v2
yet — which the deployment notes already require, but the consequence of
getting that order wrong was worse than "reports do not send".

Opting in to v2 against a v1-only collector left the installation stuck.
Registration was refused, so nothing existed at the collector at all; the row
sat in activation_pending, disable moved it to deletion_pending, retry was
futile forever, and enable refused because the row was not `disabled`. The
abandon hatch added earlier did not apply: it was gated on
SIGNING_KEY_UNREADABLE. So the most harmless possible failure — nothing
registered anywhere — was the one an operator could not clear.

The gate is now the property that actually matters: a participation the
collector has provably never accepted (sequence 0, no receipt) with a failing
delivery can be discarded, from activation_pending as well as
deletion_pending. Its receipt records `never-registered` rather than an
unconfirmed deletion, because nothing remote exists to be unsure about. A
participation the collector *did* accept keeps the old narrow gate and its
explicit warning — clearing local state while the collector still holds
reports must stay a deliberate, warned-about act.

A collector that rejects a registration or a deletion outright now reports
SCHEMA_NOT_ACCEPTED instead of DELIVERY_FAILED, and the settings page says the
collector does not accept this report version yet. Retrying cannot fix that,
and sending the operator to look for a network fault they do not have was
wrong.

Verified end to end against the live collector: v2 opt-in reports
SCHEMA_NOT_ACCEPTED, the exit is offered immediately, the receipt says
never-registered, and joining again on v1 registers, reports and withdraws
with a collector-confirmed deletion.
This commit is contained in:
Paul Nothaft
2026-09-06 18:54:58 +02:00
parent c741dc22c5
commit e40bc474bc
7 changed files with 186 additions and 22 deletions
@@ -260,3 +260,100 @@ describe('delivery backoff', () => {
expect(Number(cleared.next_attempt_at)).toBe(0);
});
});
/**
* The same dead end, reached the ordinary way. If an installation opts in to
* usage.v2 while the collector still only speaks usage.v1 — the deployment
* order the docs warn about — the registration is rejected outright. Nothing
* exists at the collector, and yet the operator could not clear the tab:
* disable moved to deletion_pending, retry was futile, enable refused, and the
* abandon hatch was gated on SIGNING_KEY_UNREADABLE, which this is not.
*
* Verified against the live collector before this was written: a valid v2
* register is answered with INVALID_PACKET while the identical v1 flow is
* accepted.
*/
describe('a participation the collector never accepted', () => {
let db;
afterEach(async () => { if (db) await db.destroy(); db = null; });
const rejectingCollector = async (status) => {
db = await bootDb();
await db.schema.createTable('product_usage_markers', (t) => {
t.string('feature', 60).primary();
});
const identity = generateIdentity();
const service = new UsageService(db, {
secret: SECRET_A,
endpoint: 'https://usage.example.test',
bindingPath: `${require('os').tmpdir()}/usage-unreg-${Date.now()}-${Math.random()}.key`,
fetch: async () => ({
ok: false,
status: 400,
headers: { get: () => null },
body: (async function* () { yield Buffer.from(JSON.stringify({ error: 'INVALID_PACKET' })); })(),
}),
});
await db('product_usage_state').where({ id: 1 }).update({
status,
consent_version: 'usage-consent.v2',
installation_id: identity.installation_id,
public_key: identity.public_key,
private_key_encrypted: service.encrypt(identity.private_key),
sequence: 0,
pending_packet: JSON.stringify(
makePacket(identity, status === 'deletion_pending' ? 'delete' : 'register', 0,
status === 'deletion_pending' ? {} : { consent_version: 'usage-consent.v2' }, 'usage.v2')
),
});
return service;
};
it('names the rejection instead of blaming the network', async () => {
const service = await rejectingCollector('activation_pending');
await service.tick({ force: true });
expect((await service.state()).last_error).toBe('SCHEMA_NOT_ACCEPTED');
});
it('offers the exit straight from activation_pending', async () => {
const service = await rejectingCollector('activation_pending');
await service.tick({ force: true });
const status = await service.status();
expect(status.can_abandon).toBe(true);
expect(status.abandon_never_registered).toBe(true);
await service.abandon();
const after = await service.state();
expect(after.status).toBe('disabled');
expect(after.installation_id).toBeNull();
// Provably nothing remote, so the receipt must not hedge.
expect(JSON.parse(after.privacy_receipts).last_abandonment.status)
.toBe('never-registered');
});
it('offers it from deletion_pending too, once the withdrawal is also undeliverable', async () => {
const service = await rejectingCollector('deletion_pending');
await service.tick({ force: true });
expect((await service.status()).can_abandon).toBe(true);
await service.abandon();
expect((await service.state()).status).toBe('disabled');
});
it('never offers it while a registered participation could still be deleted remotely', async () => {
const service = await rejectingCollector('deletion_pending');
// Something WAS accepted once: the collector may still hold reports, so
// clearing local state silently would be a lie.
await db('product_usage_state').where({ id: 1 }).update({
sequence: 3,
last_receipt: JSON.stringify({ status: 'accepted' }),
last_error: 'DELIVERY_FAILED',
});
expect((await service.status()).can_abandon).toBe(false);
await expect(service.abandon()).rejects.toThrow(/cannot be completed/);
});
it('does not offer it before a delivery has actually failed', async () => {
const service = await rejectingCollector('activation_pending');
expect((await service.status()).can_abandon).toBe(false);
});
});
+51 -17
View File
@@ -106,6 +106,29 @@ const parse = (value) => {
};
class UsageService {
// The collector has provably never accepted anything from this identity:
// no packet was ever acknowledged, so there is nothing remote to delete.
// Abandoning such a participation is harmless, which is why it may be
// offered without the warning the registered case needs.
static neverAccepted(state) {
return Number(state?.sequence || 0) === 0 && !state?.last_receipt;
}
// The two ways a participation can reach a state no amount of retrying will
// resolve. Kept as one predicate so the settings page and the endpoint can
// never disagree about whether the exit is available.
static abandonable(state) {
if (!['activation_pending', 'deletion_pending'].includes(state?.status))
return false;
// The signing key is gone: the delete packet can never be produced.
if (state.last_error === 'SIGNING_KEY_UNREADABLE') return true;
// Or the collector never accepted anything and a delivery is failing —
// the ordinary shape of "opted in against a collector that does not speak
// this report version yet". Nothing is registered remotely, so clearing
// the local state costs nothing and is the only way out of the tab.
return Boolean(state.last_error) && UsageService.neverAccepted(state);
}
schemaVersion(state) {
return state?.consent_version === CURRENT_CONSENT_VERSION
? CURRENT_SCHEMA_VERSION : 'usage.v1';
@@ -254,14 +277,12 @@ class UsageService {
Number(state.next_attempt_at || 0) > this.now()
? Number(state.next_attempt_at)
: null,
// The one failure the operator cannot retry their way out of: the
// signing key is unreadable, so the delete packet can never be signed.
// Without this flag the settings page has no way to offer the only
// remaining exit (abandon), and the install sits in deletion_pending
// forever.
can_abandon:
state.status === 'deletion_pending' &&
state.last_error === 'SIGNING_KEY_UNREADABLE',
// Whether the only remaining exit should be offered. Without it the tab
// shows a permanent error and no control that can clear it.
can_abandon: UsageService.abandonable(state),
// Distinguishes the harmless case (nothing was ever registered, so
// abandoning deletes nothing remote) from the one that needs a warning.
abandon_never_registered: UsageService.neverAccepted(state),
pending_action: state.pending_packet
? JSON.parse(state.pending_packet).action
: null,
@@ -433,13 +454,11 @@ class UsageService {
// unconfirmed deletion is its own decision, not a side effect of opting in.
async abandon() {
await this.locked(async (state) => {
if (
state.status !== 'deletion_pending' ||
state.last_error !== 'SIGNING_KEY_UNREADABLE'
)
if (!UsageService.abandonable(state))
throw new ConflictError(
'Only an unsignable withdrawal can be abandoned'
'Only a participation that cannot be completed can be abandoned'
);
const neverRegistered = UsageService.neverAccepted(state);
await fs.unlink(this.bindingPath).catch((error) => {
if (error.code !== 'ENOENT') throw error;
});
@@ -448,7 +467,8 @@ class UsageService {
? JSON.parse(state.privacy_receipts)
: {};
await this.db('product_usage_state')
.where({ id: 1, status: 'deletion_pending' })
.where({ id: 1 })
.whereIn('status', ['activation_pending', 'deletion_pending'])
.update({
status: 'disabled',
installation_id: null,
@@ -471,8 +491,15 @@ class UsageService {
kind: 'abandonment',
receipt_id: crypto.randomUUID(),
confirmed_at: new Date(this.now()).toISOString(),
status: 'collector-unconfirmed',
reason: 'SIGNING_KEY_UNREADABLE',
// Two different truths, and the receipt has to tell them apart.
// Nothing was ever accepted -> there is provably nothing at the
// collector. Otherwise the collector may still hold reports and
// was never told to delete them; saying "unconfirmed" is the
// only honest wording for that.
status: neverRegistered
? 'never-registered'
: 'collector-unconfirmed',
reason: state.last_error,
installation_id: state.installation_id,
scope: ['local identity', 'local markers', 'local key material']
}
@@ -681,11 +708,18 @@ class UsageService {
'NOT_REGISTERED',
'PACKET_CONFLICT'
].includes(error.code);
// A collector that answers INVALID_PACKET to a registration or a deletion
// does not understand the wire version we speak — most often because it
// has not been upgraded to usage.v2 yet. Retrying cannot fix that, and
// reporting it as DELIVERY_FAILED sent the operator looking for a
// network problem they do not have.
const code = conflict
? error.code
: error.code === 'SIGNING_KEY_UNREADABLE'
? 'SIGNING_KEY_UNREADABLE'
: 'DELIVERY_FAILED';
: rejected && ['register', 'delete'].includes(packet.action)
? 'SCHEMA_NOT_ACCEPTED'
: 'DELIVERY_FAILED';
await this.db('product_usage_state')
.where({ id: 1 })
.update({ last_error: code });