Merge pull request 'performance improvements' (#342) from lx-perf-improvements into main

Performance improvements included in this PR:

- [x] Use `Bytes` at a few places where appropriate, instead of `Vec<u8>`, to reduce the number of copies
  - [x] StreamChunker now accumulates incoming slices in a `Vec<Bytes>` instead of a `VecDeque<u8>`. Replaces calls to `.extend()` and `.drain()` that were quite costly by a simple `concat()` on a vec of slices which is much more optimized
- [x] Hashing (b2, sha256, md5) is now done on a Tokio thread dedicated to cpu-intensive tasks, using `spawn_blocking`
- [x] Block manager now uses 256 independant locks instead of one big lock for writing, reduces contention when writing several/many objects in parallel
- [x] Better LMDB defaults: we now put flags `NoSync` and `NoMetaSync` to avoid `fsync` at each transaction (extremely slow). Also increased number of LMDB readers to accomodate more intensive workloads

Other changes included in this PR:

- [x] Update to hashing and MAC crates: md5 and sha2 from 0.9 to 0.10, hmac from 0.10 to 0.12
- [x] switch to `tracing_subscriber` for logs, which allows to have timing of each event

Reviewed-on: https://git.deuxfleurs.fr/Deuxfleurs/garage/pulls/342

2 files changed, 34 insertions, 17 deletions

diff --git a/src/block/block.rs b/src/block/block.rs
index 4d3fbcb8..f17bd2c0 100644
--- a/src/block/block.rs
+++ b/src/block/block.rs
@@ -1,3 +1,4 @@
+use bytes::Bytes;
 use serde::{Deserialize, Serialize};
 use zstd::stream::{decode_all as zstd_decode, Encoder};
 
@@ -61,13 +62,17 @@ impl DataBlock {
 		}
 	}
 
-	pub fn from_buffer(data: Vec<u8>, level: Option<i32>) -> DataBlock {
-		if let Some(level) = level {
-			if let Ok(data) = zstd_encode(&data[..], level) {
-				return DataBlock::Compressed(data);
+	pub async fn from_buffer(data: Bytes, level: Option<i32>) -> DataBlock {
+		tokio::task::spawn_blocking(move || {
+			if let Some(level) = level {
+				if let Ok(data) = zstd_encode(&data[..], level) {
+					return DataBlock::Compressed(data);
+				}
 			}
-		}
-		DataBlock::Plain(data)
+			DataBlock::Plain(data.to_vec()) // TODO: remove to_vec here
+		})
+		.await
+		.unwrap()
 	}
 }
 
diff --git a/src/block/manager.rs b/src/block/manager.rs
index 9240db25..b5199b62 100644
--- a/src/block/manager.rs
+++ b/src/block/manager.rs
@@ -3,12 +3,18 @@ use std::sync::Arc;
 use std::time::Duration;
 
 use async_trait::async_trait;
+use bytes::Bytes;
 use serde::{Deserialize, Serialize};
 
 use tokio::fs;
 use tokio::io::{AsyncReadExt, AsyncWriteExt};
 use tokio::sync::{mpsc, Mutex};
 
+use opentelemetry::{
+	trace::{FutureExt as OtelFutureExt, TraceContextExt, Tracer},
+	Context,
+};
+
 use garage_db as db;
 
 use garage_util::data::*;
@@ -68,7 +74,7 @@ pub struct BlockManager {
 
 	compression_level: Option<i32>,
 
-	mutation_lock: Mutex<BlockManagerLocked>,
+	mutation_lock: [Mutex<BlockManagerLocked>; 256],
 
 	pub(crate) rc: BlockRc,
 	pub resync: BlockResyncManager,
@@ -105,8 +111,6 @@ impl BlockManager {
 			.netapp
 			.endpoint("garage_block/manager.rs/Rpc".to_string());
 
-		let manager_locked = BlockManagerLocked();
-
 		let metrics = BlockManagerMetrics::new(resync.queue.clone(), resync.errors.clone());
 
 		let (scrub_tx, scrub_rx) = mpsc::channel(1);
@@ -115,7 +119,7 @@ impl BlockManager {
 			replication,
 			data_dir,
 			compression_level,
-			mutation_lock: Mutex::new(manager_locked),
+			mutation_lock: [(); 256].map(|_| Mutex::new(BlockManagerLocked())),
 			rc,
 			resync,
 			system,
@@ -174,14 +178,15 @@ impl BlockManager {
 	}
 
 	/// Send block to nodes that should have it
-	pub async fn rpc_put_block(&self, hash: Hash, data: Vec<u8>) -> Result<(), Error> {
+	pub async fn rpc_put_block(&self, hash: Hash, data: Bytes) -> Result<(), Error> {
 		let who = self.replication.write_nodes(&hash);
-		let data = DataBlock::from_buffer(data, self.compression_level);
+		let data = DataBlock::from_buffer(data, self.compression_level).await;
 		self.system
 			.rpc
 			.try_call_many(
 				&self.endpoint,
 				&who[..],
+				// TODO: remove to_vec() here
 				BlockRpc::PutBlock { hash, data },
 				RequestStrategy::with_priority(PRIO_NORMAL)
 					.with_quorum(self.replication.write_quorum())
@@ -260,14 +265,21 @@ impl BlockManager {
 		hash: &Hash,
 		data: &DataBlock,
 	) -> Result<BlockRpc, Error> {
+		let tracer = opentelemetry::global::tracer("garage");
+
 		let write_size = data.inner_buffer().len() as u64;
 
-		let res = self
-			.mutation_lock
+		let res = self.mutation_lock[hash.as_slice()[0] as usize]
 			.lock()
+			.with_context(Context::current_with_span(
+				tracer.start("Acquire mutation_lock"),
+			))
 			.await
 			.write_block(hash, data, self)
 			.bound_record_duration(&self.metrics.block_write_duration)
+			.with_context(Context::current_with_span(
+				tracer.start("BlockManagerLocked::write_block"),
+			))
 			.await?;
 
 		self.metrics.bytes_written.add(write_size);
@@ -317,7 +329,7 @@ impl BlockManager {
 		if data.verify(*hash).is_err() {
 			self.metrics.corruption_counter.add(1);
 
-			self.mutation_lock
+			self.mutation_lock[hash.as_slice()[0] as usize]
 				.lock()
 				.await
 				.move_block_to_corrupted(hash, self)
@@ -331,7 +343,7 @@ impl BlockManager {
 
 	/// Check if this node has a block and whether it needs it
 	pub(crate) async fn check_block_status(&self, hash: &Hash) -> Result<BlockStatus, Error> {
-		self.mutation_lock
+		self.mutation_lock[hash.as_slice()[0] as usize]
 			.lock()
 			.await
 			.check_block_status(hash, self)
@@ -346,7 +358,7 @@ impl BlockManager {
 
 	/// Delete block if it is not needed anymore
 	pub(crate) async fn delete_if_unneeded(&self, hash: &Hash) -> Result<(), Error> {
-		self.mutation_lock
+		self.mutation_lock[hash.as_slice()[0] as usize]
 			.lock()
 			.await
 			.delete_if_unneeded(hash, self)