Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Jump to bottom

enhancement(vector source): implement client connection limits for grpc server #21072

Closed
wants to merge 1 commit into from
Closed

enhancement(vector source): implement client connection limits for grpc server #21072

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
1 commit
Select commit
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
2 changes: 2 additions & 0 deletions src/components/validation/runner/io.rs
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -164,6 +164,8 @@ pub fn spawn_grpc_server<S>(

let server = run_grpc_server(
listen_addr.as_socket_addr(),
None,
None,
tls_settings,
service,
shutdown_signal,
Expand Down
110 changes: 110 additions & 0 deletions src/sources/util/grpc/connectionlimit.rs
View file
Open in desktop
Original file line number Diff line number Diff line change
@@ -0,0 +1,110 @@
use std::{

Check failure on line 1 in src/sources/util/grpc/connectionlimit.rs

View workflow job for this annotation

GitHub Actions / Check Spelling 

`connectionlimit` is not a recognized word. (check-file-path)
sync::{Arc, Mutex},
time::{Duration, Instant},
task::{Context, Poll},
};

use http::{Request, Response};
use hyper::Body;
use std::future::Future;
use std::pin::Pin;

use tonic::{
Status,
body::BoxBody,
transport::server::NamedService,
};
use tower::{Layer, Service};

/// A service that tracks the number of requests and elapsed time,
/// shutting down the connection gracefully if the configured limits are reached.
#[derive(Clone)]
pub struct ConnectionLimit<S> {
inner: S,
request_count: Arc<Mutex<usize>>,
Copy link
Member

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

I think an atomic could be used here instead: https://doc.rust-lang.org/std/sync/atomic/struct.AtomicUsize.html

max_requests: usize,
max_duration: Duration,
start_time: Instant,
}

impl<S> ConnectionLimit<S> {
pub fn new(inner: S, max_requests: usize, max_duration: Duration) -> Self {
Self {
inner,
request_count: Arc::new(Mutex::new(0)),
max_requests: max_requests,
max_duration: max_duration,
start_time: Instant::now(),
Copy link
Member

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Nit: I think it might be better to initialize the start_time in the call method since this is closer to when it would be "running".

}
}
}

impl<S> Service<Request<Body>> for ConnectionLimit<S>
Copy link
Member

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

I think this would end up applying one duration and request limit across all clients. In an ideal world, I think we would apply the limits per-client (that is per socket).

where
S: Service<Request<Body>, Response = Response<BoxBody>, Error = tonic::Status>
+ NamedService
+ Clone
+ Send
+ 'static,
S::Future: Send + 'static,
{
type Response = Response<BoxBody>;
type Error = S::Error;
type Future = Pin<Box<dyn Future<Output = Result<Self::Response, Self::Error>> + Send>>;

fn poll_ready(&mut self, cx: &mut Context<'_>) -> Poll<Result<(), Self::Error>> {
self.inner.poll_ready(cx)
}

fn call(&mut self, req: Request<Body>) -> Self::Future {
let max_requests = self.max_requests;
let max_duration = self.max_duration;
let request_count = Arc::clone(&self.request_count);
let start_time = self.start_time;

let elapsed_time = start_time.elapsed();

let future = self.inner.call(req);

Box::pin(async move {
let response = future.await?;

// After processing the request, increment the request count and check the limits.
let mut count = request_count.lock().unwrap();
*count += 1;

if *count > max_requests || elapsed_time > max_duration {
// If the limit is reached, return a ResourceExhausted error to close the connection.
return Err(Status::resource_exhausted(
"Connection closed after reaching the limit.",
));
Comment on lines +78 to +80
Copy link
Member

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Will the server close the connection if this error is returned? Or are we depending on the client closing it when it sees this error?

}

Ok(response)
})
}
}

/// A layer that adds the ConnectionLimit functionality to a service.
#[derive(Clone, Default)]
pub struct ConnectionLimitLayer {
max_requests: usize,
max_duration: Duration,
}

impl ConnectionLimitLayer {
pub fn new(max_requests: usize, max_duration: Duration) -> Self {
Self {
max_requests,
max_duration,
}
}
}

impl<S> Layer<S> for ConnectionLimitLayer {
type Service = ConnectionLimit<S>;

fn layer(&self, inner: S) -> Self::Service {
ConnectionLimit::new(inner, self.max_requests, self.max_duration)
}
}
15 changes: 12 additions & 3 deletions src/sources/util/grpc/mod.rs
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -6,13 +6,14 @@
use futures::FutureExt;
use http::{Request, Response};
use hyper::Body;
use std::{convert::Infallible, net::SocketAddr, time::Duration};
use std::{net::SocketAddr, time::Duration};
use tonic::transport::server::Routes;
use tonic::{
Status,
body::BoxBody,
transport::server::{NamedService, Server},
};
use tower::Service;
use tower::{Layer, Service};
use tower_http::{
classify::{GrpcErrorsAsFailures, SharedClassifier},
trace::TraceLayer,
Expand All @@ -22,14 +23,19 @@
mod decompression;
pub use self::decompression::{DecompressionAndMetrics, DecompressionAndMetricsLayer};

mod connectionlimit;

Check failure on line 26 in src/sources/util/grpc/mod.rs

View workflow job for this annotation

GitHub Actions / Check Spelling 

`connectionlimit` is not a recognized word. (unrecognized-spelling)
Copy link
Member

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Nit: module names usually use _ for delimiting multiple words so this would be connection_limit.

pub use self::connectionlimit::{ConnectionLimit, ConnectionLimitLayer};

Check failure on line 27 in src/sources/util/grpc/mod.rs

View workflow job for this annotation

GitHub Actions / Check Spelling 

`connectionlimit` is not a recognized word. (unrecognized-spelling)

pub async fn run_grpc_server<S>(
address: SocketAddr,
max_requests: usize,
max_duration: Duration,
tls_settings: MaybeTlsSettings,
service: S,
shutdown: ShutdownSignal,
) -> crate::Result<()>
where
S: Service<Request<Body>, Response = Response<BoxBody>, Error = Infallible>
S: Service<Request<Body>, Response = Response<BoxBody>, Error = tonic::Status>
+ NamedService
+ Clone
+ Send
Expand All @@ -43,6 +49,9 @@

info!(%address, "Building gRPC server.");

// Conditionally apply the ConnectionLimitLayer if any limits are set
let service = ConnectionLimitLayer::new(max_requests, max_duration).layer(service);
Comment on lines +52 to +53
Copy link
Member

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

I think this could be added in the list before as another layer after (or before) the DecompressionAndMetricsLayer.


Server::builder()
.layer(build_grpc_trace_layer(span.clone()))
// This layer explicitly decompresses payloads, if compressed, and reports the number of message bytes we've
Expand Down
22 changes: 20 additions & 2 deletions src/sources/vector/mod.rs
View file
Open in desktop
Original file line number Diff line number Diff line change
@@ -1,5 +1,8 @@
//! The `vector` source. See [VectorConfig].
use std::net::SocketAddr;
use std::{
net::SocketAddr,
time::Duration,
};

use chrono::Utc;
use futures::TryFutureExt;
Expand All @@ -26,6 +29,8 @@ use crate::{
SourceSender,
};

use serde_with::serde_as;

/// Marker type for version two of the configuration for the `vector` source.
#[configurable_component]
#[derive(Clone, Debug)]
Expand Down Expand Up @@ -114,6 +119,7 @@ async fn handle_batch_status(receiver: Option<BatchStatusReceiver>) -> Result<()
}

/// Configuration for the `vector` source.
#[serde_as]
#[configurable_component(source("vector", "Collect observability data from a Vector instance."))]
#[derive(Clone, Debug)]
#[serde(deny_unknown_fields)]
Expand All @@ -134,6 +140,16 @@ pub struct VectorConfig {
#[serde(default, deserialize_with = "bool_or_struct")]
acknowledgements: SourceAcknowledgementsConfig,

/// Maximum duration of client connection before it is closed
#[serde_as(as = "serde_with::DurationSeconds<u64>")]
#[configurable(metadata(docs::human_name = "Max client connection duration"))]
max_duration: Duration,

/// Maximum number of client requests before connection is closed
#[configurable(metadata(docs::type_unit = "requests"))]
#[configurable(metadata(docs::human_name = "Max client requests before connection is closed"))]
max_requests: usize,

/// The namespace to use for logs. This overrides the global setting.
#[serde(default)]
#[configurable(metadata(docs::hidden))]
Expand All @@ -157,6 +173,8 @@ impl Default for VectorConfig {
address: "0.0.0.0:6000".parse().unwrap(),
tls: None,
acknowledgements: Default::default(),
max_requests: usize::MAX,
max_duration: Duration::from_secs(u64::MAX),
log_namespace: None,
}
}
Expand Down Expand Up @@ -186,7 +204,7 @@ impl SourceConfig for VectorConfig {
.max_decoding_message_size(usize::MAX);

let source =
run_grpc_server(self.address, tls_settings, service, cx.shutdown).map_err(|error| {
run_grpc_server(self.address, self.max_requests, self.max_duration, tls_settings, service, cx.shutdown).map_err(|error| {
error!(message = "Source future failed.", %error);
});

Expand Down
Loading