Re-hash keys before storing in LMDB to improve key distribution and avoid prefixes

doc/book/build/python.md

@@ -23,7 +23,7 @@ client = minio.Minio(
   "GKyourapikey",
   "abcd[...]1234",
   # Force the region, this is specific to garage
-  region="garage",
+  region="region",
 )
 ```
 

doc/book/connect/apps/index.md

@@ -335,7 +335,6 @@ From the [official Mastodon documentation](https://docs.joinmastodon.org/admin/t
 
 ```bash
 $ RAILS_ENV=production bin/tootctl media remove --days 3
-$ RAILS_ENV=production bin/tootctl media remove --days 15 --prune-profiles
 $ RAILS_ENV=production bin/tootctl media remove-orphans
 $ RAILS_ENV=production bin/tootctl preview_cards remove --days 15
 ```
@@ -354,6 +353,8 @@ Imports:	1.7 KB
 Settings:	0 Bytes
 ```
 
+Unfortunately, [old avatars and headers cannot currently be cleaned up](https://github.com/mastodon/mastodon/issues/9567).
+
 ### Migrating your data
 
 Data migration should be done with an efficient S3 client.

doc/book/design/goals.md

@@ -50,20 +50,3 @@ locations. They use Garage themselves for the following tasks:
 
 The Deuxfleurs Garage cluster is a multi-site cluster currently composed of
 9 nodes in 3 physical locations.
-
-### Triplebit
-
-[Triplebit](https://www.triplebit.org) is a non-profit hosting provider and
-ISP focused on improving access to privacy-related services. They use
-Garage themselves for the following tasks:
-
-- Hosting of their homepage, [privacyguides.org](https://www.privacyguides.org/), and various other static sites
-
-- As a Mastodon object storage backend for [mstdn.party](https://mstdn.party/) and [mstdn.plus](https://mstdn.plus/)
-
-- As a PeerTube storage backend for [neat.tube](https://neat.tube/)
-
-- As a [Matrix media backend](https://github.com/matrix-org/synapse-s3-storage-provider)
-
-Triplebit's Garage cluster is a multi-site cluster currently composed of
-10 nodes in 3 physical locations.

doc/book/reference-manual/configuration.md

@@ -16,7 +16,6 @@ data_dir = "/var/lib/garage/data"
 metadata_fsync = true
 data_fsync = false
 disable_scrub = false
-use_local_tz = false
 metadata_auto_snapshot_interval = "6h"
 
 db_engine = "lmdb"
@@ -100,7 +99,6 @@ Top-level configuration options:
 [`data_fsync`](#data_fsync),
 [`db_engine`](#db_engine),
 [`disable_scrub`](#disable_scrub),
-[`use_local_tz`](#use_local_tz),
 [`lmdb_map_size`](#lmdb_map_size),
 [`metadata_auto_snapshot_interval`](#metadata_auto_snapshot_interval),
 [`metadata_dir`](#metadata_dir),
@@ -429,13 +427,6 @@ you should delete it from the data directory and then call `garage repair
 blocks` on the node to ensure that it re-obtains a copy from another node on
 the network.
 
-#### `use_local_tz` {#use_local_tz}
-
-By default, Garage runs the lifecycle worker every day at midnight in UTC. Set the
-`use_local_tz` configuration value to `true` if you want Garage to run the
-lifecycle worker at midnight in your local timezone. If you have multiple nodes,
-you should also ensure that each node has the same timezone configuration.
-
 #### `block_size` {#block_size}
 
 Garage splits stored objects in consecutive chunks of size `block_size`

script/helm/garage/templates/workload.yaml

@@ -76,9 +76,6 @@ spec:
             - name: etc
               mountPath: /etc/garage.toml
               subPath: garage.toml
-          {{- with .Values.extraVolumeMounts }}
-            {{- toYaml . | nindent 12 }}
-          {{- end }}
           # TODO
           # livenessProbe:
           #   httpGet:
@@ -113,9 +110,6 @@ spec:
         - name: data
           emptyDir: {}
         {{- end }}
-        {{- with .Values.extraVolumes }}
-          {{- toYaml . | nindent 8 }}
-        {{- end }}
       {{- with .Values.nodeSelector }}
       nodeSelector:
         {{- toYaml . | nindent 8 }}

script/helm/garage/values.yaml

@@ -218,10 +218,6 @@ affinity: {}
 
 environment: {}
 
-extraVolumes: {}
-
-extraVolumeMounts: {}
-
 monitoring:
   metrics:
     # If true, a service for monitoring is created with a prometheus.io/scrape annotation

src/db/Cargo.toml

@@ -15,6 +15,9 @@ path = "lib.rs"
 err-derive.workspace = true
 hexdump.workspace = true
 tracing.workspace = true
+opentelemetry.workspace = true
+opentelemetry.workspace = true
+xxhash-rust.workspace = true
 
 heed = { workspace = true, optional = true }
 rusqlite = { workspace = true, optional = true, features = ["backup"] }

src/db/lmdb_adapter.rs

@@ -10,6 +10,8 @@ use std::sync::{Arc, RwLock};
 use heed::types::ByteSlice;
 use heed::{BytesDecode, Env, RoTxn, RwTxn, UntypedDatabase as Database};
 
+use xxhash_rust::xxh3::xxh3_128;
+
 use crate::{
 	Db, Error, IDb, ITx, ITxFn, OnCommit, Result, TxError, TxFnResult, TxOpError, TxOpResult,
 	TxResult, TxValueIter, Value, ValueIter,
@@ -58,6 +60,40 @@ impl LmdbDb {
 	}
 }
 
+fn key_hash(key: &[u8]) -> [u8; 16] {
+    xxh3_128(key).to_ne_bytes()
+}
+
+fn kv_to_value(key: &[u8], value: &[u8]) -> Vec<u8> {
+    [&key.len().to_ne_bytes(), key, value].concat()
+}
+
+fn value_to_kv(value: &[u8]) -> (Vec<u8>, Vec<u8>) {
+    const USIZE_LEN: usize = std::mem::size_of::<usize>();
+    let klen = usize::from_ne_bytes(value[0..USIZE_LEN].try_into().unwrap());
+    (
+        value[USIZE_LEN..klen+USIZE_LEN].to_vec(),
+        value[USIZE_LEN+klen..].to_vec()
+    )
+}
+
+fn key_hash(key: &[u8]) -> [u8; 16] {
+    xxh3_128(key).to_ne_bytes()
+}
+
+fn kv_to_value(key: &[u8], value: &[u8]) -> Vec<u8> {
+    [&key.len().to_ne_bytes(), key, value].concat()
+}
+
+fn value_to_kv(value: &[u8]) -> (Vec<u8>, Vec<u8>) {
+    const USIZE_LEN: usize = std::mem::size_of::<usize>();
+    let klen = usize::from_ne_bytes(value[0..USIZE_LEN].try_into().unwrap());
+    (
+        value[USIZE_LEN..klen+USIZE_LEN].to_vec(),
+        value[USIZE_LEN+klen..].to_vec()
+    )
+}
+
 impl IDb for LmdbDb {
 	fn engine(&self) -> String {
 		"LMDB (using Heed crate)".into()
@@ -119,10 +155,11 @@ impl IDb for LmdbDb {
 		let tree = self.get_tree(tree)?;
 
 		let tx = self.db.read_txn()?;
-		let val = tree.get(&tx, key)?;
+        let kh = key_hash(key);
+		let val = tree.get(&tx, &kh)?;
 		match val {
 			None => Ok(None),
-			Some(v) => Ok(Some(v.to_vec())),
+			Some(v) => Ok(Some(value_to_kv(v).1))
 		}
 	}
 
@@ -135,7 +172,9 @@ impl IDb for LmdbDb {
 	fn insert(&self, tree: usize, key: &[u8], value: &[u8]) -> Result<()> {
 		let tree = self.get_tree(tree)?;
 		let mut tx = self.db.write_txn()?;
-		tree.put(&mut tx, key, value)?;
+        let kh = key_hash(key);
+        let value = kv_to_value(key, value);
+		tree.put(&mut tx, &kh, &value)?;
 		tx.commit()?;
 		Ok(())
 	}
@@ -143,7 +182,8 @@ impl IDb for LmdbDb {
 	fn remove(&self, tree: usize, key: &[u8]) -> Result<()> {
 		let tree = self.get_tree(tree)?;
 		let mut tx = self.db.write_txn()?;
-		tree.delete(&mut tx, key)?;
+        let kh = key_hash(key);
+		tree.delete(&mut tx, &kh)?;
 		tx.commit()?;
 		Ok(())
 	}
@@ -242,8 +282,9 @@ impl<'a> LmdbTx<'a> {
 impl<'a> ITx for LmdbTx<'a> {
 	fn get(&self, tree: usize, key: &[u8]) -> TxOpResult<Option<Value>> {
 		let tree = self.get_tree(tree)?;
-		match tree.get(&self.tx, key)? {
-			Some(v) => Ok(Some(v.to_vec())),
+        let kh = key_hash(key);
+		match tree.get(&self.tx, &kh)? {
+			Some(v) => Ok(Some(value_to_kv(v).1)),
 			None => Ok(None),
 		}
 	}
@@ -254,14 +295,18 @@ impl<'a> ITx for LmdbTx<'a> {
 
 	fn insert(&mut self, tree: usize, key: &[u8], value: &[u8]) -> TxOpResult<()> {
 		let tree = *self.get_tree(tree)?;
-		tree.put(&mut self.tx, key, value)?;
+        let kh = key_hash(key);
+        let value = kv_to_value(key, value);
+		tree.put(&mut self.tx, &kh, &value)?;
 		Ok(())
 	}
 	fn remove(&mut self, tree: usize, key: &[u8]) -> TxOpResult<()> {
 		let tree = *self.get_tree(tree)?;
-		tree.delete(&mut self.tx, key)?;
+        let kh = key_hash(key);
+		tree.delete(&mut self.tx, &kh)?;
 		Ok(())
 	}
+
 	fn clear(&mut self, tree: usize) -> TxOpResult<()> {
 		let tree = *self.get_tree(tree)?;
 		tree.clear(&mut self.tx)?;
@@ -370,7 +415,7 @@ where
 		match next {
 			None => None,
 			Some(Err(e)) => Some(Err(e.into())),
-			Some(Ok((k, v))) => Some(Ok((k.to_vec(), v.to_vec()))),
+			Some(Ok((_k, v))) => Some(Ok(value_to_kv(v))),
 		}
 	}
 }
@@ -380,7 +425,7 @@ where
 fn tx_iter_item<'a>(
 	item: std::result::Result<(&'a [u8], &'a [u8]), heed::Error>,
 ) -> TxOpResult<(Vec<u8>, Vec<u8>)> {
-	item.map(|(k, v)| (k.to_vec(), v.to_vec()))
+	item.map(|(_k, v)| value_to_kv(v))
 		.map_err(|e| TxOpError(Error::from(e)))
 }
 

src/model/garage.rs

@@ -141,7 +141,7 @@ impl Garage {
 		)?)
 		.ok()
 		.and_then(|x| NetworkKey::from_slice(&x))
-		.ok_or_message("Invalid RPC secret key: expected 32 bytes of random hex, please check the documentation for requirements")?;
+		.ok_or_message("Invalid RPC secret key: expected 32 bits of entropy, please check the documentation for requirements")?;
 
 		let (replication_factor, consistency_mode) = parse_replication_mode(&config)?;
 

src/model/s3/lifecycle_worker.rs

@@ -70,7 +70,7 @@ pub fn register_bg_vars(
 
 impl LifecycleWorker {
 	pub fn new(garage: Arc<Garage>, persister: PersisterShared<LifecycleWorkerPersisted>) -> Self {
-		let today = today(garage.config.use_local_tz);
+		let today = today();
 		let last_completed = persister.get_with(|x| {
 			x.last_completed
 				.as_deref()
@@ -205,9 +205,8 @@ impl Worker for LifecycleWorker {
 	async fn wait_for_work(&mut self) -> WorkerState {
 		match &self.state {
 			State::Completed(d) => {
-				let use_local_tz = self.garage.config.use_local_tz;
 				let next_day = d.succ_opt().expect("no next day");
-				let next_start = midnight_ts(next_day, use_local_tz);
+				let next_start = midnight_ts(next_day);
 				loop {
 					let now = now_msec();
 					if now < next_start {
@@ -219,7 +218,7 @@ impl Worker for LifecycleWorker {
 						break;
 					}
 				}
-				self.state = State::start(std::cmp::max(next_day, today(use_local_tz)));
+				self.state = State::start(std::cmp::max(next_day, today()));
 			}
 			State::Running { .. } => (),
 		}
@@ -386,16 +385,10 @@ fn check_size_filter(version_data: &ObjectVersionData, filter: &LifecycleFilter)
 	true
 }
 
-fn midnight_ts(date: NaiveDate, use_local_tz: bool) -> u64 {
-	let midnight = date.and_hms_opt(0, 0, 0).expect("midnight does not exist");
-	if use_local_tz {
-		return midnight
-			.and_local_timezone(Local)
-			.single()
-			.expect("bad local midnight")
-			.timestamp_millis() as u64;
-	}
-	midnight.timestamp_millis() as u64
+fn midnight_ts(date: NaiveDate) -> u64 {
+	date.and_hms_opt(0, 0, 0)
+		.expect("midnight does not exist")
+		.timestamp_millis() as u64
 }
 
 fn next_date(ts: u64) -> NaiveDate {
@@ -406,9 +399,6 @@ fn next_date(ts: u64) -> NaiveDate {
 		.expect("no next day")
 }
 
-fn today(use_local_tz: bool) -> NaiveDate {
-	if use_local_tz {
-		return Local::now().naive_local().date();
-	}
+fn today() -> NaiveDate {
 	Utc::now().naive_utc().date()
 }

src/util/config.rs

@@ -27,10 +27,6 @@ pub struct Config {
 	#[serde(default)]
 	pub disable_scrub: bool,
 
-	/// Use local timezone
-	#[serde(default)]
-	pub use_local_tz: bool,
-
 	/// Automatic snapshot interval for metadata
 	#[serde(default)]
 	pub metadata_auto_snapshot_interval: Option<String>,