Compare commits

...
Author SHA1 Message Date
Brandon Thomas 12269f8bfb flags 2026-03-25 23:37:00 -04:00
Brandon Thomas bf02bf7bbc nice control flags 2026-03-25 23:35:00 -04:00
Brandon Thomas 14ef9cb2b8 break apart into easier to manage pieces 2026-03-25 22:31:02 -04:00
Brandon Thomas 50da7c66b6 alerting middleware with error types 2026-03-25 22:21:17 -04:00
Brandon Thomas 25a4169c6b error reporting middleware 2026-03-25 21:27:39 -04:00
13 changed files with 312 additions and 5 deletions
@@ -8,6 +8,10 @@ COOKIE_SECURE=false
COOKIE_DOMAIN=.jungle.horse
WEBSITE_HOMEPAGE_REDIRECT=http://jungle.horse
# Paging
ENABLE_PAGING=true
ENABLE_PAGING_FOR_500S=true
# Test the rate limiter bypass (obviously these are dummy values)
AI_STREAMER_USERNAMES="PewDiePie,xQc"
@@ -0,0 +1,36 @@
use pager::client::pager::Pager;
use crate::http_server::common_responses::common_web_error::CommonWebError;
use super::enqueue_alert::enqueue_alert;
/// Check `CommonWebError` and alert on server errors.
/// Returns `true` if the error was handled (alerted or intentionally skipped).
pub(super) fn check_common_web_error(
pager: &Pager,
method: &str,
path: &str,
error: &CommonWebError,
) -> bool {
match error {
CommonWebError::ServerError => {
enqueue_alert(
pager,
format!("CommonWebError::ServerError on {} {}", method, path),
format!(
"A CommonWebError::ServerError was returned.\n\n\
Endpoint: {} {}\n\
Error: {:?}\n\
Time: {}",
method, path, error,
chrono::Utc::now().format("%Y-%m-%d %H:%M:%S UTC"),
),
);
true
}
// Don't alert on client errors (400, 401, 404, 402).
CommonWebError::BadInputWithSimpleMessage(_) => true,
CommonWebError::NotFound => true,
CommonWebError::NotAuthorized => true,
CommonWebError::PaymentRequired => true,
}
}
@@ -0,0 +1,30 @@
use pager::client::pager::Pager;
use super::enqueue_alert::enqueue_alert;
/// Fallback alerting based on HTTP status code when no typed error matched.
pub(super) fn check_status_code_fallback(
pager: &Pager,
method: &str,
path: &str,
status_code: u16,
) {
match status_code {
500 => {
enqueue_alert(
pager,
format!("HTTP 500: {} {}", method, path),
format!(
"An untyped HTTP 500 response was returned (no typed error matched).\n\n\
Endpoint: {} {}\n\
Time: {}",
method, path,
chrono::Utc::now().format("%Y-%m-%d %H:%M:%S UTC"),
),
);
}
// Add more status-based rules here:
// 503 => { enqueue_alert(...); }
_ => {}
}
}
@@ -0,0 +1,14 @@
use log::{debug, warn};
use pager::client::pager::Pager;
use pager::notification::notification_details::NotificationDetails;
pub(super) fn enqueue_alert(pager: &Pager, summary: String, description: String) {
let notification = NotificationDetails::with_summary_and_description(summary, description);
if let Err(err) = pager.enqueue_page(notification) {
warn!("Error alerting middleware: failed to enqueue page: {:?}", err);
} else {
debug!("Error alerting middleware: enqueued alert");
}
}
@@ -0,0 +1,153 @@
use std::future::{Future, Ready, ready};
use std::pin::Pin;
use actix_web::dev::{Service, ServiceRequest, ServiceResponse, Transform};
use actix_web::Error;
use pager::client::pager::Pager;
use crate::http_server::common_responses::common_web_error::CommonWebError;
use crate::state::flags::paging_flags::PagingFlags;
use super::check_common_web_error::check_common_web_error;
use super::check_status_code_fallback::check_status_code_fallback;
// ======================== Transform (factory) ========================
/// Middleware that intercepts error responses and enqueues pager alerts.
///
/// Inspects errors in two ways:
/// 1. **Typed error matching** via downcast (e.g. `CommonWebError::ServerError`)
/// 2. **Status code fallback** for untyped 500s that slip through
///
/// Both `PagingFlags.is_paging_enabled` and `PagingFlags.is_paging_for_500s_enabled`
/// must be true for any alerting to fire. Otherwise the middleware is a passthrough.
///
/// To add new error types, create a new `check_*.rs` matcher module and
/// add a downcast branch in `check_ok_response_for_alerts` / `check_err_for_alerts`.
#[derive(Clone)]
pub struct ErrorAlertingMiddleware {
pager: Pager,
paging_flags: PagingFlags,
}
impl ErrorAlertingMiddleware {
pub fn new(pager: Pager, paging_flags: PagingFlags) -> Self {
Self { pager, paging_flags }
}
}
impl<S, B> Transform<S, ServiceRequest> for ErrorAlertingMiddleware
where
S: Service<ServiceRequest, Response = ServiceResponse<B>, Error = Error> + 'static,
S::Future: 'static,
B: 'static,
{
type Response = ServiceResponse<B>;
type Error = Error;
type InitError = ();
type Transform = ErrorAlertingService<S>;
type Future = Ready<Result<Self::Transform, Self::InitError>>;
fn new_transform(&self, service: S) -> Self::Future {
ready(Ok(ErrorAlertingService {
service,
pager: self.pager.clone(),
paging_flags: self.paging_flags.clone(),
}))
}
}
// ======================== Service (per-request) ========================
pub struct ErrorAlertingService<S> {
service: S,
pager: Pager,
paging_flags: PagingFlags,
}
impl<S, B> Service<ServiceRequest> for ErrorAlertingService<S>
where
S: Service<ServiceRequest, Response = ServiceResponse<B>, Error = Error> + 'static,
S::Future: 'static,
B: 'static,
{
type Response = ServiceResponse<B>;
type Error = Error;
type Future = Pin<Box<dyn Future<Output = Result<Self::Response, Self::Error>>>>;
actix_service::forward_ready!(service);
fn call(&self, req: ServiceRequest) -> Self::Future {
// Fast path: if paging is disabled, skip all inspection.
if !self.paging_flags.is_paging_enabled || !self.paging_flags.is_paging_for_500s_enabled {
return Box::pin(self.service.call(req));
}
let method = req.method().to_string();
let path = req.path().to_string();
let pager = self.pager.clone();
let fut = self.service.call(req);
Box::pin(async move {
match fut.await {
Ok(res) => {
check_ok_response_for_alerts(&pager, &method, &path, &res);
Ok(res)
}
Err(err) => {
check_err_for_alerts(&pager, &method, &path, &err);
Err(err)
}
}
})
}
}
// ======================== Response inspection ========================
/// Inspect a successful `ServiceResponse` that may contain an error response.
///
/// Actix-web converts `ResponseError` types into HTTP responses and stashes
/// the original error in `response.error()`. We can downcast from there.
fn check_ok_response_for_alerts<B>(
pager: &Pager,
method: &str,
path: &str,
response: &ServiceResponse<B>,
) {
let status = response.status();
// Try to get the original typed error from the response (if it came from ResponseError).
if let Some(err) = response.response().error() {
// --- Typed matchers (add new check_*.rs modules and downcast branches here) ---
if let Some(common_err) = err.as_error::<CommonWebError>() {
if check_common_web_error(pager, method, path, common_err) {
return;
}
}
}
// --- Status code fallback ---
check_status_code_fallback(pager, method, path, status.as_u16());
}
/// Inspect an `actix_web::Error` returned from the handler or inner middleware.
fn check_err_for_alerts(
pager: &Pager,
method: &str,
path: &str,
err: &Error,
) {
// --- Typed matchers (add new check_*.rs modules and downcast branches here) ---
if let Some(common_err) = err.as_error::<CommonWebError>() {
if check_common_web_error(pager, method, path, common_err) {
return;
}
}
// --- Status code fallback ---
let status = err.as_response_error().status_code();
check_status_code_fallback(pager, method, path, status.as_u16());
}
@@ -0,0 +1,5 @@
mod check_common_web_error;
mod check_status_code_fallback;
mod enqueue_alert;
pub mod error_alerting_middleware;
@@ -1 +1,2 @@
pub mod error_alerting_middleware;
pub mod pushback_filter_middleware;
+18 -1
View File
@@ -72,6 +72,7 @@ use crate::configs::app_startup::redis_rate_limiters::configure_redis_rate_limit
use crate::configs::connect_to_database::connect_to_database;
use crate::configs::static_api_tokens::StaticApiTokenSet;
use crate::http_server::cookies::anonymous_visitor_tracking::avt_cookie_manager::AvtCookieManager;
use crate::http_server::middleware::error_alerting_middleware::error_alerting_middleware::ErrorAlertingMiddleware;
use crate::http_server::middleware::pushback_filter_middleware::PushbackFilter;
use crate::http_server::routes::add_routes::add_routes;
use crate::http_server::session::session_checker::SessionChecker;
@@ -338,7 +339,7 @@ async fn main() -> AnyhowResult<()> {
ServerEnvironment::Development => server_environment::ServerEnvironment::Development,
};
let (pager, pager_worker) = build_pager(server_environment_typed);
let (pager, pager_worker, paging_flags) = build_pager(server_environment_typed);
info!("Spawning pager worker thread.");
@@ -391,6 +392,9 @@ async fn main() -> AnyhowResult<()> {
// Disable voice features
disable_tts: easyenv::get_env_bool_or_default("DISABLE_TTS", false),
disable_voice_conversion: easyenv::get_env_bool_or_default("DISABLE_VOICE_CONVERSION", false),
// Paging
paging: paging_flags,
};
let third_party_url_redirector = ThirdPartyUrlRedirector::new(server_environment);
@@ -625,6 +629,15 @@ pub async fn serve(server_state: ServerState) -> AnyhowResult<()>
ServerEnvironment::Production => server_environment::ServerEnvironment::Production,
};
let paging_flags_for_middleware = server_state.flags.paging.clone();
let pager_for_middleware = server_state.pager.clone();
let enable_error_alerting = paging_flags_for_middleware.is_paging_enabled
&& paging_flags_for_middleware.is_paging_for_500s_enabled;
if enable_error_alerting {
info!("Error alerting middleware is ENABLED (ENABLE_PAGING=true, ENABLE_PAGING_FOR_500S=true).");
}
let server_state_arc = web::Data::new(Arc::new(server_state));
let disabled_endpoints = read_disabled_endpoints();
@@ -682,6 +695,10 @@ pub async fn serve(server_state: ServerState) -> AnyhowResult<()>
.wrap(DefaultHeaders::new()
.header("X-Backend-Hostname", &hostname)
.header("X-Build-Sha", server_state_arc.server_info.build_sha.clone()))
.wrap(middleware::Condition::new(
enable_error_alerting,
ErrorAlertingMiddleware::new(pager_for_middleware.clone(), paging_flags_for_middleware.clone()),
))
.wrap(PushbackFilter::new(&server_state_arc.flags.clone()))
.wrap(DisabledEndpointFilter::new(disabled_endpoints.clone()))
.wrap(BannedIpFilter::new(ip_ban_list))
@@ -4,20 +4,38 @@ use pager::client::pager_builder::PagerBuilder;
use pager::worker::pager_worker::PagerWorker;
use rootly_client::creds::rootly_api_key::RootlyApiKey;
pub fn build_pager(server_environment: server_environment::ServerEnvironment) -> (Pager, PagerWorker) {
use crate::state::flags::paging_flags::PagingFlags;
pub fn build_pager(
server_environment: server_environment::ServerEnvironment,
) -> (Pager, PagerWorker, PagingFlags) {
let paging_flags = PagingFlags::from_env();
info!("Paging flags: {:?}", paging_flags);
let environment = if server_environment.is_deployed_in_production() {
"production"
} else {
"development"
};
// If paging is globally disabled, use a NoOp pager regardless of API key.
if !paging_flags.is_paging_enabled {
warn!("ENABLE_PAGING is false. Pager will be NoOp.");
let (pager, worker) = PagerBuilder::new()
.application_name("storyteller-web".to_string())
.environment(environment.to_string())
.build_with_worker();
return (pager, worker, paging_flags);
}
let maybe_api_key = easyenv::get_env_string_optional("ROOTLY_API_KEY");
let builder = PagerBuilder::new()
.application_name("storyteller-web".to_string())
.environment(environment.to_string());
match maybe_api_key {
let (pager, worker) = match maybe_api_key {
Some(api_key) => {
info!("Rootly API key found. Configuring pager with Rootly backend.");
build_rootly_pager(builder, api_key)
@@ -26,10 +44,12 @@ pub fn build_pager(server_environment: server_environment::ServerEnvironment) ->
warn!("ROOTLY_API_KEY not set. Pager will not send real pages.");
builder.build_with_worker()
}
}
};
(pager, worker, paging_flags)
}
fn build_rootly_pager(mut builder: PagerBuilder, api_key: String) -> (Pager, PagerWorker) {
fn build_rootly_pager(builder: PagerBuilder, api_key: String) -> (Pager, PagerWorker) {
let mut rootly_builder = builder
.rootly(RootlyApiKey::new(api_key));
@@ -0,0 +1 @@
pub mod paging_flags;
@@ -0,0 +1,21 @@
/// Flags controlling the pager system.
#[derive(Clone, Debug)]
pub struct PagingFlags {
/// Master switch for paging. If false, no pages are sent.
/// Env var: ENABLE_PAGING (default: false)
pub is_paging_enabled: bool,
/// If true, the error alerting middleware will enqueue pages for HTTP 500s.
/// Requires `is_paging_enabled` to also be true.
/// Env var: ENABLE_PAGING_FOR_500S (default: false)
pub is_paging_for_500s_enabled: bool,
}
impl PagingFlags {
pub fn from_env() -> Self {
Self {
is_paging_enabled: easyenv::get_env_bool_or_default("ENABLE_PAGING", false),
is_paging_for_500s_enabled: easyenv::get_env_bool_or_default("ENABLE_PAGING_FOR_500S", false),
}
}
}
@@ -1,4 +1,5 @@
pub mod cached_queries;
pub mod flags;
pub mod certs;
pub mod memory_cache;
pub mod server_state;
@@ -33,6 +33,7 @@ use crate::http_server::endpoints::stats::result_transformer::CacheableQueueStat
use crate::http_server::endpoints::tts::list_tts_models::TtsModelRecordForResponse;
use crate::http_server::endpoints::voice_conversion::list_voice_conversion_models_handler::VoiceConversionModel;
use crate::http_server::session::session_checker::SessionChecker;
use crate::state::flags::paging_flags::PagingFlags;
use crate::http_server::web_utils::redis_rate_limiter::RedisRateLimiter;
use crate::http_server::web_utils::scoped_temp_dir_creator::ScopedTempDirCreator;
use crate::state::certs::google_sign_in_cert::GoogleSignInCert;
@@ -298,6 +299,9 @@ pub struct StaticFeatureFlags {
/// Disable voice conversion endpoints
/// If true, respond with 429.
pub disable_voice_conversion: bool,
/// Paging system flags.
pub paging: PagingFlags,
}
/// Instead of top level service denial, these are bans against entities that instead return