connlib: different backoff strategy for gateway/client (#1910)

With this PR we will keep retrying reconnection forever for the gateway
after it disconnects.

rust/Cargo.lock

@@ -369,9 +369,12 @@ version = "0.4.0"
 source = "registry+https://github.com/rust-lang/crates.io-index"
 checksum = "b62ddb9cb1ec0a098ad4bbf9344d0713fa193ae1a80af55febcff2627b6a00c1"
 dependencies = [
+ "futures-core",
  "getrandom",
  "instant",
+ "pin-project-lite",
  "rand",
+ "tokio",
 ]
 
 [[package]]
@@ -1115,6 +1118,7 @@ name = "firezone-client-connlib"
 version = "0.1.0"
 dependencies = [
  "async-trait",
+ "backoff",
  "boringtun",
  "chrono",
  "firezone-tunnel",
@@ -1130,6 +1134,7 @@ name = "firezone-gateway-connlib"
 version = "0.1.0"
 dependencies = [
  "async-trait",
+ "backoff",
  "boringtun",
  "chrono",
  "firezone-tunnel",

rust/Cargo.toml

@@ -16,6 +16,7 @@ members = [
 boringtun = { git = "https://github.com/firezone/boringtun", branch = "master", default-features = false }
 chrono = { version = "0.4", default-features = false, features = ["std", "clock", "oldtime", "serde"] }
 swift-bridge = "0.1.52"
+backoff = { version = "0.4", features = ["tokio"] }
 
 # Patched to use https://github.com/rust-lang/cc-rs/pull/708
 # (the `patch` section can't be used for build deps...)

rust/connlib/libs/client/Cargo.toml

@@ -14,6 +14,7 @@ libs-common = { path = "../common" }
 firezone-tunnel = { path = "../tunnel" }
 serde = { version = "1.0", default-features = false, features = ["std", "derive"] }
 boringtun = { workspace = true }
+backoff = { workspace = true }
 
 [dev-dependencies]
 serde_json = { version = "1.0", default-features = false, features = ["std"] }

rust/connlib/libs/client/src/control.rs

@@ -1,6 +1,7 @@
 use std::{sync::Arc, time::Duration};
 
 use crate::messages::{Connect, ConnectionDetails, EgressMessages, InitClient, Messages};
+use backoff::{ExponentialBackoff, ExponentialBackoffBuilder};
 use boringtun::x25519::StaticSecret;
 use libs_common::{
     control::{ErrorInfo, ErrorReply, MessageResult, PhoenixSenderWithTopic},
@@ -254,4 +255,8 @@ impl<CB: Callbacks + 'static> ControlSession<Messages, CB> for ControlPlane<CB>
     fn socket_path() -> &'static str {
         "device"
     }
+
+    fn retry_strategy() -> ExponentialBackoff {
+        ExponentialBackoffBuilder::default().build()
+    }
 }

rust/connlib/libs/common/Cargo.toml

@@ -23,7 +23,7 @@ tokio = { version = "1.28", default-features = false, features = ["rt", "rt-mult
 url = { version = "2.3.1", default-features = false }
 rand_core = { version = "0.6.4", default-features = false, features = ["std"] }
 async-trait = { version = "0.1", default-features = false }
-backoff = { version = "0.4", default-features = false }
+backoff = { workspace = true }
 ip_network = { version = "0.4", default-features = false, features = ["serde"] }
 boringtun = { workspace = true }
 os_info = { version = "3", default-features = false }

rust/connlib/libs/common/src/session.rs

@@ -1,5 +1,5 @@
 use async_trait::async_trait;
-use backoff::{backoff::Backoff, ExponentialBackoffBuilder};
+use backoff::{backoff::Backoff, ExponentialBackoff};
 use boringtun::x25519::{PublicKey, StaticSecret};
 use ip_network::IpNetwork;
 use parking_lot::Mutex;
@@ -40,6 +40,9 @@ pub trait ControlSession<T, CB: Callbacks> {
 
     /// Either "gateway" or "client" used to get the control-plane URL.
     fn socket_path() -> &'static str;
+
+    /// Retry strategy in case of disconnection for the session.
+    fn retry_strategy() -> ExponentialBackoff;
 }
 
 // TODO: Currently I'm using Session for both gateway and clients
@@ -306,7 +309,7 @@ where
             );
 
             tokio::spawn(async move {
-                let mut exponential_backoff = ExponentialBackoffBuilder::default().build();
+                let mut exponential_backoff = T::retry_strategy();
                 loop {
                     // `connection.start` calls the callback only after connecting
                     let result = connection.start(vec![topic.clone()], || exponential_backoff.reset()).await;

rust/connlib/libs/gateway/Cargo.toml

@@ -12,6 +12,7 @@ tracing = { version = "0.1", default-features = false, features = ["std", "attri
 serde = { version = "1.0", default-features = false, features = ["std", "derive"] }
 boringtun = { workspace = true }
 chrono = { workspace = true }
+backoff = { workspace = true }
 
 [dev-dependencies]
 serde_json = { version = "1.0", default-features = false, features = ["std"] }

rust/connlib/libs/gateway/src/control.rs

@@ -1,5 +1,6 @@
 use std::{sync::Arc, time::Duration};
 
+use backoff::{ExponentialBackoff, ExponentialBackoffBuilder};
 use boringtun::x25519::StaticSecret;
 use firezone_tunnel::{ControlSignal, Tunnel};
 use libs_common::{
@@ -164,4 +165,10 @@ impl<CB: Callbacks + 'static> ControlSession<IngressMessages, CB> for ControlPla
     fn socket_path() -> &'static str {
         "gateway"
     }
+
+    fn retry_strategy() -> ExponentialBackoff {
+        ExponentialBackoffBuilder::default()
+            .with_max_elapsed_time(None)
+            .build()
+    }
 }