Webhook DLQ — list failed deliveries and manually retry

Closes the silent-loss hole in outbound webhook delivery. The worker
in src/webhooks.rs retries failed deliveries with exponential backoff
up to 10 attempts, then sets next_attempt_at = NULL and walks away.
Pre-this-commit, those "dead-lettered" rows sat in webhook_deliveries
forever with no surface for the operator to discover, inspect, or
recover from them — a subscriber that was down for >6h during a
license-issuance burst would silently lose those events forever.

What's new:

- repo::DeliveryStatusFilter — enum with parse() so query strings
  map cleanly to SQL predicates.
- repo::list_deliveries — endpoint_id + status + limit, newest first.
- repo::requeue_delivery — resets attempt_count=0, clears delivered_at
  and last_error, sets next_attempt_at=now. The worker picks it up on
  the next 5s tick.

- src/api/webhook_deliveries.rs — admin module with two handlers:
  - GET /v1/admin/webhook-deliveries?endpoint_id=…&status=…&limit=…
  - POST /v1/admin/webhook-deliveries/:id/retry  (audit-logged as
    webhook_delivery.retry; 404 on missing id)
- Routes registered in src/api/mod.rs alongside the existing
  webhook_endpoints CRUD.

- tests/api.rs gains webhook_dlq_lists_failed_and_retry_requeues:
  seeds three deliveries directly via SQL (one each: delivered,
  pending, dead-lettered), exercises the list filter, runs the retry,
  asserts the row migrates from failed→pending, audit row is written,
  404 on bad id, 400 on bad status filter.

Worker code is unchanged. The DLQ is operator-actionable infrastructure
on top of the existing retry semantics.

Test count: 23 (9 unit + 4 migration + 10 API), up from 22.
This commit is contained in:
Grant
2026-05-08 09:38:58 -05:00
parent e2b296ce29
commit f9ef1a854c
4 changed files with 387 additions and 0 deletions
+11
View File
@@ -70,6 +70,7 @@ pub mod session_layer;
pub mod tier;
pub mod validate;
pub mod webhook;
pub mod webhook_deliveries;
pub mod webhook_endpoints;
use crate::btcpay::client::BtcpayClient;
@@ -304,6 +305,16 @@ pub fn router(state: AppState) -> Router {
"/v1/admin/webhook-endpoints/:id",
axum::routing::delete(webhook_endpoints::delete),
)
// Webhook delivery history (the dead-letter inspection +
// manual-retry surface; see webhook_deliveries.rs for why).
.route(
"/v1/admin/webhook-deliveries",
get(webhook_deliveries::list),
)
.route(
"/v1/admin/webhook-deliveries/:id/retry",
post(webhook_deliveries::retry),
)
// Discount / referral codes.
.route(
"/v1/admin/discount-codes",
@@ -0,0 +1,104 @@
//! Admin views over the outbound webhook delivery queue.
//!
//! Companion to `webhook_endpoints.rs`: that module manages the
//! configured subscriber URLs; this one exposes the row-level history
//! of attempts (success, in-flight retries, dead-lettered failures)
//! and lets operators manually re-queue a dead delivery for another
//! pass through the worker.
//!
//! Why this exists: the worker in `crate::webhooks` retries failed
//! deliveries with exponential backoff up to 10 attempts, then sets
//! `next_attempt_at = NULL` and walks away. Pre-this-module, those
//! "dead-lettered" rows were invisible — operators had no surface to
//! discover, inspect, or recover from them. A subscriber endpoint
//! that was down for >6h during a license-issuance burst would
//! silently lose those events forever.
use crate::api::admin::{request_context, require_admin};
use crate::api::AppState;
use crate::db::repo::{self, DeliveryStatusFilter};
use crate::error::{AppError, AppResult};
use axum::{
extract::{Path, Query, State},
http::HeaderMap,
Json,
};
use serde::Deserialize;
use serde_json::{json, Value};
const DEFAULT_LIMIT: i64 = 100;
const MAX_LIMIT: i64 = 500;
#[derive(Debug, Deserialize)]
pub struct ListDeliveriesQuery {
/// Filter by configured endpoint id. Omit for all endpoints.
pub endpoint_id: Option<String>,
/// One of `pending` | `delivered` | `failed` | `all`. Defaults to
/// `all`. The `failed` filter is the dead-letter queue — rows
/// where the worker exhausted retries.
pub status: Option<String>,
/// Cap on rows returned. Defaults to 100; max 500.
pub limit: Option<i64>,
}
pub async fn list(
State(state): State<AppState>,
headers: HeaderMap,
Query(q): Query<ListDeliveriesQuery>,
) -> AppResult<Json<Value>> {
require_admin(&state, &headers)?;
let status = match q.status.as_deref() {
Some(s) => DeliveryStatusFilter::parse(s).ok_or_else(|| {
AppError::BadRequest(format!(
"invalid status filter '{s}'; expected pending|delivered|failed|all"
))
})?,
None => DeliveryStatusFilter::All,
};
let limit = q
.limit
.unwrap_or(DEFAULT_LIMIT)
.clamp(1, MAX_LIMIT);
let rows = repo::list_deliveries(
&state.db,
q.endpoint_id.as_deref(),
status,
limit,
)
.await?;
Ok(Json(json!({ "deliveries": rows })))
}
/// Manual re-queue for a dead-lettered (or otherwise stuck)
/// delivery. The worker will pick it up on the next 5s tick.
///
/// 404 if the delivery id doesn't exist; 200 on success with the
/// updated row in the body so the SPA can re-render the list with
/// the new state immediately.
pub async fn retry(
State(state): State<AppState>,
headers: HeaderMap,
Path(id): Path<String>,
) -> AppResult<Json<Value>> {
let actor_hash = require_admin(&state, &headers)?;
let (ip, ua) = request_context(&headers);
let delivery = repo::requeue_delivery(&state.db, &id)
.await?
.ok_or_else(|| AppError::NotFound(format!("webhook delivery '{id}'")))?;
let _ = repo::insert_audit(
&state.db,
"admin_api_key",
Some(&actor_hash),
"webhook_delivery.retry",
Some("webhook_delivery"),
Some(&id),
ip.as_deref(),
ua.as_deref(),
&json!({
"endpoint_id": delivery.endpoint_id,
"event_type": delivery.event_type,
}),
)
.await;
Ok(Json(json!(delivery)))
}
+108
View File
@@ -1379,6 +1379,114 @@ pub async fn mark_delivery_failure(
Ok(())
}
/// Filter modes for `list_deliveries`. Strings match the values
/// accepted by the admin endpoint's `?status=...` query param.
pub enum DeliveryStatusFilter {
/// `delivered_at IS NULL AND next_attempt_at IS NOT NULL` — in
/// the retry queue, will be picked up by the worker on the next
/// tick that's past `next_attempt_at`.
Pending,
/// `delivered_at IS NOT NULL` — successfully delivered.
Delivered,
/// `delivered_at IS NULL AND next_attempt_at IS NULL AND
/// attempt_count > 0` — the dead-letter case. Worker exhausted
/// retries (or hit a hard error like a deleted endpoint) and
/// won't re-pick it. Operators see these via the admin list and
/// can manually re-queue via `requeue_delivery`.
Failed,
/// All deliveries.
All,
}
impl DeliveryStatusFilter {
pub fn parse(s: &str) -> Option<Self> {
match s {
"pending" => Some(Self::Pending),
"delivered" => Some(Self::Delivered),
"failed" => Some(Self::Failed),
"all" => Some(Self::All),
_ => None,
}
}
}
/// List webhook deliveries with optional filtering. Newest first
/// (orders by `created_at DESC`) so the admin UI shows recent
/// activity at the top.
pub async fn list_deliveries(
pool: &SqlitePool,
endpoint_id: Option<&str>,
status: DeliveryStatusFilter,
limit: i64,
) -> AppResult<Vec<WebhookDelivery>> {
let mut sql = String::from(
"SELECT id, endpoint_id, event_type, payload_json, attempt_count,
next_attempt_at, last_status_code, last_error, delivered_at, created_at
FROM webhook_deliveries WHERE 1=1",
);
if endpoint_id.is_some() {
sql.push_str(" AND endpoint_id = ?");
}
match status {
DeliveryStatusFilter::Pending => {
sql.push_str(" AND delivered_at IS NULL AND next_attempt_at IS NOT NULL")
}
DeliveryStatusFilter::Delivered => sql.push_str(" AND delivered_at IS NOT NULL"),
DeliveryStatusFilter::Failed => sql.push_str(
" AND delivered_at IS NULL AND next_attempt_at IS NULL AND attempt_count > 0",
),
DeliveryStatusFilter::All => {}
}
sql.push_str(" ORDER BY created_at DESC LIMIT ?");
let mut q = sqlx::query(&sql);
if let Some(eid) = endpoint_id {
q = q.bind(eid);
}
q = q.bind(limit);
let rows = q.fetch_all(pool).await?;
Ok(rows.into_iter().map(row_to_delivery).collect())
}
/// Re-queue a previously-failed (or even successfully-delivered)
/// delivery for another attempt. Resets `attempt_count` to 0, clears
/// `delivered_at` and `last_error`, and sets `next_attempt_at` to
/// now so the worker picks it up on the next tick.
///
/// Returns the affected row, or `Ok(None)` if no row with the given
/// id exists.
pub async fn requeue_delivery(
pool: &SqlitePool,
id: &str,
) -> AppResult<Option<WebhookDelivery>> {
let now = Utc::now().to_rfc3339();
let res = sqlx::query(
"UPDATE webhook_deliveries
SET attempt_count = 0,
delivered_at = NULL,
last_error = NULL,
last_status_code = NULL,
next_attempt_at = ?
WHERE id = ?",
)
.bind(&now)
.bind(id)
.execute(pool)
.await?;
if res.rows_affected() == 0 {
return Ok(None);
}
let row = sqlx::query(
"SELECT id, endpoint_id, event_type, payload_json, attempt_count,
next_attempt_at, last_status_code, last_error, delivered_at, created_at
FROM webhook_deliveries WHERE id = ?",
)
.bind(id)
.fetch_one(pool)
.await?;
Ok(Some(row_to_delivery(row)))
}
fn row_to_delivery(row: sqlx::sqlite::SqliteRow) -> WebhookDelivery {
WebhookDelivery {
id: row.get("id"),
+164
View File
@@ -827,3 +827,167 @@ async fn tier_caps_block_at_creator_limit_and_unlock_after_upgrade() {
.unwrap();
assert_eq!(count, 6, "the previously-blocked product should now exist");
}
/// Webhook DLQ (dead-letter queue) — list + retry round trip.
///
/// The delivery worker retries failed deliveries with exponential
/// backoff up to 10 attempts, then sets `next_attempt_at = NULL` and
/// walks away. Pre-this-feature, those rows were invisible to the
/// operator. Now `GET /v1/admin/webhook-deliveries?status=failed`
/// surfaces them and `POST /v1/admin/webhook-deliveries/:id/retry`
/// puts them back in the queue.
///
/// We seed a "dead-lettered" row directly via SQL — the worker isn't
/// spawned in tests, so we don't need to drive 10 real failures to
/// reach the dead state. This tests the admin surface, not the
/// worker.
#[tokio::test]
async fn webhook_dlq_lists_failed_and_retry_requeues() {
let (state, _tmp) = make_test_state().await;
let auth = format!("Bearer {}", TEST_ADMIN_KEY);
let now = Utc::now().to_rfc3339();
// Configure a webhook endpoint to own the deliveries.
let endpoint_id = "ep1";
sqlx::query(
"INSERT INTO webhook_endpoints(id, url, secret, event_types, active, \
description, created_at, updated_at) \
VALUES(?, 'https://operator.example/keysat-hook', \
'0123456789abcdef0123456789abcdef', '[\"*\"]', 1, '', ?, ?)",
)
.bind(endpoint_id)
.bind(&now)
.bind(&now)
.execute(&state.db)
.await
.unwrap();
// One delivery in each state: delivered (success), pending
// (in-queue), and failed (DLQ — what we mostly care about).
let mk = |id: &str, attempts: i64, next: Option<&str>, delivered: Option<&str>| {
let id = id.to_string();
let attempts = attempts;
let next = next.map(|s| s.to_string());
let delivered = delivered.map(|s| s.to_string());
let pool = state.db.clone();
let now = now.clone();
async move {
sqlx::query(
"INSERT INTO webhook_deliveries(id, endpoint_id, event_type, \
payload_json, attempt_count, next_attempt_at, delivered_at, created_at) \
VALUES(?, ?, 'license.issued', '{}', ?, ?, ?, ?)",
)
.bind(&id)
.bind(endpoint_id)
.bind(attempts)
.bind(next.as_deref())
.bind(delivered.as_deref())
.bind(&now)
.execute(&pool)
.await
.unwrap();
}
};
mk("d-delivered", 1, None, Some(&now)).await;
mk("d-pending", 2, Some(&now), None).await;
// The dead-lettered case: 10 attempts, next_attempt_at NULL, never delivered.
mk("d-failed", 10, None, None).await;
// List with status=failed should return ONLY the dead-lettered row.
let req = build_request(
"GET",
"/v1/admin/webhook-deliveries?status=failed",
&[("authorization", &auth)],
None,
);
let resp = send(&state, req).await;
assert_eq!(resp.status(), StatusCode::OK);
let body = body_json(resp).await;
let deliveries = body["deliveries"].as_array().expect("deliveries array");
assert_eq!(
deliveries.len(),
1,
"status=failed should return the one DLQ row, got {deliveries:?}"
);
assert_eq!(deliveries[0]["id"], "d-failed");
assert_eq!(deliveries[0]["attempt_count"], 10);
// Retry the dead-lettered delivery.
let req = build_request(
"POST",
"/v1/admin/webhook-deliveries/d-failed/retry",
&[("authorization", &auth)],
None,
);
let resp = send(&state, req).await;
assert_eq!(resp.status(), StatusCode::OK, "retry should succeed");
let body = body_json(resp).await;
assert_eq!(
body["attempt_count"], 0,
"retry should reset attempt_count to 0"
);
assert!(
body["next_attempt_at"].is_string(),
"retry should set next_attempt_at: {body:?}"
);
// After retry: status=failed should be empty (the row left the
// DLQ); status=pending should now contain it.
let req = build_request(
"GET",
"/v1/admin/webhook-deliveries?status=failed",
&[("authorization", &auth)],
None,
);
let resp = send(&state, req).await;
let body = body_json(resp).await;
assert_eq!(
body["deliveries"].as_array().unwrap().len(),
0,
"after retry, the row should no longer be 'failed'"
);
let req = build_request(
"GET",
"/v1/admin/webhook-deliveries?status=pending",
&[("authorization", &auth)],
None,
);
let resp = send(&state, req).await;
let body = body_json(resp).await;
let pending = body["deliveries"].as_array().unwrap();
assert!(
pending.iter().any(|d| d["id"] == "d-failed"),
"after retry, the previously-failed row should appear in 'pending'"
);
// Audit log captured the retry.
let audit_count: i64 = sqlx::query_scalar(
"SELECT COUNT(*) FROM audit_log WHERE action = 'webhook_delivery.retry' AND target_id = 'd-failed'",
)
.fetch_one(&state.db)
.await
.unwrap();
assert_eq!(audit_count, 1, "retry must write an audit log entry");
// Retry on a non-existent id is 404.
let req = build_request(
"POST",
"/v1/admin/webhook-deliveries/never-existed/retry",
&[("authorization", &auth)],
None,
);
let resp = send(&state, req).await;
assert_eq!(resp.status(), StatusCode::NOT_FOUND);
// Bad status filter is 400 (a typo'd query string shouldn't
// silently succeed; that's a UI footgun).
let req = build_request(
"GET",
"/v1/admin/webhook-deliveries?status=garbage",
&[("authorization", &auth)],
None,
);
let resp = send(&state, req).await;
assert_eq!(resp.status(), StatusCode::BAD_REQUEST);
}