mirror of
https://github.com/huggingface/text-generation-inference.git
synced 2025-04-20 14:22:08 +00:00
feat(server): add local prom and health routes if running w/ ngrok
This commit is contained in:
parent
15b3e9ffb0
commit
1da642bd0e
@ -683,10 +683,10 @@ pub async fn run(
|
|||||||
// Prometheus metrics route
|
// Prometheus metrics route
|
||||||
.route("/metrics", get(metrics))
|
.route("/metrics", get(metrics))
|
||||||
.layer(Extension(info))
|
.layer(Extension(info))
|
||||||
.layer(Extension(health_ext))
|
.layer(Extension(health_ext.clone()))
|
||||||
.layer(Extension(compat_return_full_text))
|
.layer(Extension(compat_return_full_text))
|
||||||
.layer(Extension(infer))
|
.layer(Extension(infer))
|
||||||
.layer(Extension(prom_handle))
|
.layer(Extension(prom_handle.clone()))
|
||||||
.layer(opentelemetry_tracing_layer())
|
.layer(opentelemetry_tracing_layer())
|
||||||
.layer(cors_layer);
|
.layer(cors_layer);
|
||||||
|
|
||||||
@ -712,6 +712,21 @@ pub async fn run(
|
|||||||
|
|
||||||
let listener = tunnel.listen().await.unwrap();
|
let listener = tunnel.listen().await.unwrap();
|
||||||
|
|
||||||
|
// Run prom metrics and health locally too
|
||||||
|
tokio::spawn(
|
||||||
|
axum::Server::bind(&addr)
|
||||||
|
.serve(
|
||||||
|
Router::new()
|
||||||
|
.route("/health", get(health))
|
||||||
|
.route("/metrics", get(metrics))
|
||||||
|
.layer(Extension(health_ext))
|
||||||
|
.layer(Extension(prom_handle))
|
||||||
|
.into_make_service(),
|
||||||
|
)
|
||||||
|
//Wait until all requests are finished to shut down
|
||||||
|
.with_graceful_shutdown(shutdown_signal()),
|
||||||
|
);
|
||||||
|
|
||||||
// Run server
|
// Run server
|
||||||
axum::Server::builder(listener)
|
axum::Server::builder(listener)
|
||||||
.serve(app.into_make_service())
|
.serve(app.into_make_service())
|
||||||
|
Loading…
Reference in New Issue
Block a user