Proxy: remove fail fast logic to connect to compute (#6759)

## Problem

Flaky tests

## Summary of changes

Remove failfast logic

proxy/src/proxy/connect_compute.rs

@@ -122,25 +122,24 @@ where
 
     error!(error = ?err, "could not connect to compute node");
 
-    let node_info =
-        if err.get_error_kind() == crate::error::ErrorKind::Postgres || !node_info.cached() {
-            // If the error is Postgres, that means that we managed to connect to the compute node, but there was an error.
-            // Do not need to retrieve a new node_info, just return the old one.
-            if !err.should_retry(num_retries) {
-                return Err(err.into());
-            }
-            node_info
-        } else {
-            // if we failed to connect, it's likely that the compute node was suspended, wake a new compute node
-            info!("compute node's state has likely changed; requesting a wake-up");
-            ctx.latency_timer.cache_miss();
-            let old_node_info = invalidate_cache(node_info);
-            let mut node_info = wake_compute(&mut num_retries, ctx, user_info).await?;
-            node_info.reuse_settings(old_node_info);
-
-            mechanism.update_connect_config(&mut node_info.config);
-            node_info
-        };
+    let node_info = if !node_info.cached() {
+        // If we just recieved this from cplane and dodn't get it from cache, we shouldn't retry.
+        // Do not need to retrieve a new node_info, just return the old one.
+        if !err.should_retry(num_retries) {
+            return Err(err.into());
+        }
+        node_info
+    } else {
+        // if we failed to connect, it's likely that the compute node was suspended, wake a new compute node
+        info!("compute node's state has likely changed; requesting a wake-up");
+        ctx.latency_timer.cache_miss();
+        let old_node_info = invalidate_cache(node_info);
+        let mut node_info = wake_compute(&mut num_retries, ctx, user_info).await?;
+        node_info.reuse_settings(old_node_info);
+
+        mechanism.update_connect_config(&mut node_info.config);
+        node_info
+    };
 
     // now that we have a new node, try connect to it repeatedly.
     // this can error for a few reasons, for instance:

proxy/src/proxy/tests.rs

@@ -375,8 +375,6 @@ enum ConnectAction {
     Connect,
     Retry,
     Fail,
-    RetryPg,
-    FailPg,
 }
 
 #[derive(Clone)]
@@ -466,14 +464,6 @@ impl ConnectMechanism for TestConnectMechanism {
                 retryable: false,
                 kind: ErrorKind::Compute,
             }),
-            ConnectAction::FailPg => Err(TestConnectError {
-                retryable: false,
-                kind: ErrorKind::Postgres,
-            }),
-            ConnectAction::RetryPg => Err(TestConnectError {
-                retryable: true,
-                kind: ErrorKind::Postgres,
-            }),
             x => panic!("expecting action {:?}, connect is called instead", x),
         }
     }
@@ -572,32 +562,6 @@ async fn connect_to_compute_retry() {
     mechanism.verify();
 }
 
-#[tokio::test]
-async fn connect_to_compute_retry_pg() {
-    let _ = env_logger::try_init();
-    use ConnectAction::*;
-    let mut ctx = RequestMonitoring::test();
-    let mechanism = TestConnectMechanism::new(vec![Wake, RetryPg, Connect]);
-    let user_info = helper_create_connect_info(&mechanism);
-    connect_to_compute(&mut ctx, &mechanism, &user_info, false)
-        .await
-        .unwrap();
-    mechanism.verify();
-}
-
-#[tokio::test]
-async fn connect_to_compute_fail_pg() {
-    let _ = env_logger::try_init();
-    use ConnectAction::*;
-    let mut ctx = RequestMonitoring::test();
-    let mechanism = TestConnectMechanism::new(vec![Wake, FailPg]);
-    let user_info = helper_create_connect_info(&mechanism);
-    connect_to_compute(&mut ctx, &mechanism, &user_info, false)
-        .await
-        .unwrap_err();
-    mechanism.verify();
-}
-
 /// Test that we don't retry if the error is not retryable.
 #[tokio::test]
 async fn connect_to_compute_non_retry_1() {