From 1686d9e7332653ef44d8690d4764c582437eb6bf Mon Sep 17 00:00:00 2001
From: Christian Schwarz <christian@neon.tech>
Date: Thu, 6 Feb 2025 09:33:37 +0100
Subject: [PATCH] perf(page_service): dont `.instrument(span.clone())` the
 response flush (#10686)

On my AX102 Hetzner box, removing this line removes about 20us from the
`latency_mean` result in

`test_pageserver_characterize_latencies_with_1_client_and_throughput_with_many_clients_one_tenant`.

If the same 20us can be removed in the nightly benchmark run, this will
be a ~10% improvement because there, mean latencies are about ~220us.

This span was added during batching refactors, we didn't have it before,
and I don't think it's terribly useful.

refs
- https://github.com/neondatabase/cloud/issues/21759
---
 pageserver/src/page_service.rs | 2 --
 1 file changed, 2 deletions(-)

diff --git a/pageserver/src/page_service.rs b/pageserver/src/page_service.rs
index e103338c7c..679cf3b2d5 100644
--- a/pageserver/src/page_service.rs
+++ b/pageserver/src/page_service.rs
@@ -1280,8 +1280,6 @@ impl PageServerHandler {
                 }
                 Ok(())
             }
-            // and log the info! line inside the request span
-            .instrument(span.clone())
             .await?;
         }
         Ok(())