From b44d3fc796484a50cd6854f20c9b46e5fddedc9d Mon Sep 17 00:00:00 2001 From: Alex Date: Wed, 8 Jun 2022 10:01:44 +0200 Subject: Abstract database behind generic interface and implement alternative drivers (#322) - [x] Design interface - [x] Implement Sled backend - [x] Re-implement the SledCountedTree hack ~~on Sled backend~~ on all backends (i.e. over the abstraction) - [x] Convert Garage code to use generic interface - [x] Proof-read converted Garage code - [ ] Test everything well - [x] Implement sqlite backend - [x] Implement LMDB backend - [ ] (Implement Persy backend?) - [ ] (Implement other backends? (like RocksDB, ...)) - [x] Implement backend choice in config file and garage server module - [x] Add CLI for converting between DB formats - Exploit the new interface to put more things in transactions - [x] `.updated()` trigger on Garage tables Fix #284 **Bugs** - [x] When exporting sqlite, trees iterate empty?? - [x] LMDB doesn't work **Known issues for various back-ends** - Sled: - Eats all my RAM and also all my disk space - `.len()` has to traverse the whole table - Is actually quite slow on some operations - And is actually pretty bad code... - Sqlite: - Requires a lock to be taken on all operations. The lock is also taken when iterating on a table with `.iter()`, and the lock isn't released until the iterator is dropped. This means that we must be VERY carefull to not do anything else inside a `.iter()` loop or else we will have a deadlock! Most such cases have been eliminated from the Garage codebase, but there might still be some that remain. If your Garage-over-Sqlite seems to hang/freeze, this is the reason. - (adapter uses a bunch of unsafe code) - Heed (LMDB): - Not suited for 32-bit machines as it has to map the whole DB in memory. - (adpater uses a tiny bit of unsafe code) **My recommendation:** avoid 32-bit machines and use LMDB as much as possible. **Converting databases** is actually quite easy. For example from Sled to LMDB: ```bash cd src/db cargo run --features cli --bin convert -- -i path/to/garage/meta/db -a sled -o path/to/garage/meta/db.lmdb -b lmdb ``` Then, just add this to your `config.toml`: ```toml db_engine = "lmdb" ``` Co-authored-by: Alex Auvolat Reviewed-on: https://git.deuxfleurs.fr/Deuxfleurs/garage/pulls/322 Co-authored-by: Alex Co-committed-by: Alex --- src/db/sled_adapter.rs | 260 +++++++++++++++++++++++++++++++++++++++++++++++++ 1 file changed, 260 insertions(+) create mode 100644 src/db/sled_adapter.rs (limited to 'src/db/sled_adapter.rs') diff --git a/src/db/sled_adapter.rs b/src/db/sled_adapter.rs new file mode 100644 index 00000000..982f8d82 --- /dev/null +++ b/src/db/sled_adapter.rs @@ -0,0 +1,260 @@ +use core::ops::Bound; + +use std::cell::Cell; +use std::collections::HashMap; +use std::sync::{Arc, RwLock}; + +use sled::transaction::{ + ConflictableTransactionError, TransactionError, Transactional, TransactionalTree, + UnabortableTransactionError, +}; + +use crate::{ + Db, Error, IDb, ITx, ITxFn, Result, TxError, TxFnResult, TxOpError, TxOpResult, TxResult, + TxValueIter, Value, ValueIter, +}; + +pub use sled; + +// -- err + +impl From for Error { + fn from(e: sled::Error) -> Error { + Error(format!("Sled: {}", e).into()) + } +} + +impl From for TxOpError { + fn from(e: sled::Error) -> TxOpError { + TxOpError(e.into()) + } +} + +// -- db + +pub struct SledDb { + db: sled::Db, + trees: RwLock<(Vec, HashMap)>, +} + +impl SledDb { + pub fn init(db: sled::Db) -> Db { + let s = Self { + db, + trees: RwLock::new((Vec::new(), HashMap::new())), + }; + Db(Arc::new(s)) + } + + fn get_tree(&self, i: usize) -> Result { + self.trees + .read() + .unwrap() + .0 + .get(i) + .cloned() + .ok_or_else(|| Error("invalid tree id".into())) + } +} + +impl IDb for SledDb { + fn engine(&self) -> String { + "Sled".into() + } + + fn open_tree(&self, name: &str) -> Result { + let mut trees = self.trees.write().unwrap(); + if let Some(i) = trees.1.get(name) { + Ok(*i) + } else { + let tree = self.db.open_tree(name)?; + let i = trees.0.len(); + trees.0.push(tree); + trees.1.insert(name.to_string(), i); + Ok(i) + } + } + + fn list_trees(&self) -> Result> { + let mut trees = vec![]; + for name in self.db.tree_names() { + let name = std::str::from_utf8(&name) + .map_err(|e| Error(format!("{}", e).into()))? + .to_string(); + if name != "__sled__default" { + trees.push(name); + } + } + Ok(trees) + } + + // ---- + + fn get(&self, tree: usize, key: &[u8]) -> Result> { + let tree = self.get_tree(tree)?; + let val = tree.get(key)?; + Ok(val.map(|x| x.to_vec())) + } + + fn len(&self, tree: usize) -> Result { + let tree = self.get_tree(tree)?; + Ok(tree.len()) + } + + fn insert(&self, tree: usize, key: &[u8], value: &[u8]) -> Result> { + let tree = self.get_tree(tree)?; + let old_val = tree.insert(key, value)?; + Ok(old_val.map(|x| x.to_vec())) + } + + fn remove(&self, tree: usize, key: &[u8]) -> Result> { + let tree = self.get_tree(tree)?; + let old_val = tree.remove(key)?; + Ok(old_val.map(|x| x.to_vec())) + } + + fn iter(&self, tree: usize) -> Result> { + let tree = self.get_tree(tree)?; + Ok(Box::new(tree.iter().map(|v| { + v.map(|(x, y)| (x.to_vec(), y.to_vec())).map_err(Into::into) + }))) + } + + fn iter_rev(&self, tree: usize) -> Result> { + let tree = self.get_tree(tree)?; + Ok(Box::new(tree.iter().rev().map(|v| { + v.map(|(x, y)| (x.to_vec(), y.to_vec())).map_err(Into::into) + }))) + } + + fn range<'r>( + &self, + tree: usize, + low: Bound<&'r [u8]>, + high: Bound<&'r [u8]>, + ) -> Result> { + let tree = self.get_tree(tree)?; + Ok(Box::new(tree.range::<&'r [u8], _>((low, high)).map(|v| { + v.map(|(x, y)| (x.to_vec(), y.to_vec())).map_err(Into::into) + }))) + } + fn range_rev<'r>( + &self, + tree: usize, + low: Bound<&'r [u8]>, + high: Bound<&'r [u8]>, + ) -> Result> { + let tree = self.get_tree(tree)?; + Ok(Box::new(tree.range::<&'r [u8], _>((low, high)).rev().map( + |v| v.map(|(x, y)| (x.to_vec(), y.to_vec())).map_err(Into::into), + ))) + } + + // ---- + + fn transaction(&self, f: &dyn ITxFn) -> TxResult<(), ()> { + let trees = self.trees.read().unwrap(); + let res = trees.0.transaction(|txtrees| { + let mut tx = SledTx { + trees: txtrees, + err: Cell::new(None), + }; + match f.try_on(&mut tx) { + TxFnResult::Ok => { + assert!(tx.err.into_inner().is_none()); + Ok(()) + } + TxFnResult::Abort => { + assert!(tx.err.into_inner().is_none()); + Err(ConflictableTransactionError::Abort(())) + } + TxFnResult::DbErr => { + let e = tx.err.into_inner().expect("No DB error"); + Err(e.into()) + } + } + }); + match res { + Ok(()) => Ok(()), + Err(TransactionError::Abort(())) => Err(TxError::Abort(())), + Err(TransactionError::Storage(s)) => Err(TxError::Db(s.into())), + } + } +} + +// ---- + +struct SledTx<'a> { + trees: &'a [TransactionalTree], + err: Cell>, +} + +impl<'a> SledTx<'a> { + fn get_tree(&self, i: usize) -> TxOpResult<&TransactionalTree> { + self.trees.get(i).ok_or_else(|| { + TxOpError(Error( + "invalid tree id (it might have been openned after the transaction started)".into(), + )) + }) + } + + fn save_error( + &self, + v: std::result::Result, + ) -> TxOpResult { + match v { + Ok(x) => Ok(x), + Err(e) => { + let txt = format!("{}", e); + self.err.set(Some(e)); + Err(TxOpError(Error(txt.into()))) + } + } + } +} + +impl<'a> ITx for SledTx<'a> { + fn get(&self, tree: usize, key: &[u8]) -> TxOpResult> { + let tree = self.get_tree(tree)?; + let tmp = self.save_error(tree.get(key))?; + Ok(tmp.map(|x| x.to_vec())) + } + fn len(&self, _tree: usize) -> TxOpResult { + unimplemented!(".len() in transaction not supported with Sled backend") + } + + fn insert(&mut self, tree: usize, key: &[u8], value: &[u8]) -> TxOpResult> { + let tree = self.get_tree(tree)?; + let old_val = self.save_error(tree.insert(key, value))?; + Ok(old_val.map(|x| x.to_vec())) + } + fn remove(&mut self, tree: usize, key: &[u8]) -> TxOpResult> { + let tree = self.get_tree(tree)?; + let old_val = self.save_error(tree.remove(key))?; + Ok(old_val.map(|x| x.to_vec())) + } + + fn iter(&self, _tree: usize) -> TxOpResult> { + unimplemented!("Iterators in transactions not supported with Sled backend"); + } + fn iter_rev(&self, _tree: usize) -> TxOpResult> { + unimplemented!("Iterators in transactions not supported with Sled backend"); + } + + fn range<'r>( + &self, + _tree: usize, + _low: Bound<&'r [u8]>, + _high: Bound<&'r [u8]>, + ) -> TxOpResult> { + unimplemented!("Iterators in transactions not supported with Sled backend"); + } + fn range_rev<'r>( + &self, + _tree: usize, + _low: Bound<&'r [u8]>, + _high: Bound<&'r [u8]>, + ) -> TxOpResult> { + unimplemented!("Iterators in transactions not supported with Sled backend"); + } +} -- cgit v1.2.3 From 77e3fd6db2c9cd3a10889bd071e95ef839cfbefc Mon Sep 17 00:00:00 2001 From: Alex Date: Wed, 15 Jun 2022 20:20:28 +0200 Subject: improve internal item counter mechanisms and implement bucket quotas (#326) - [x] Refactoring of internal counting API - [x] Repair procedure for counters (it's an offline procedure!!!) - [x] New counter for objects in buckets - [x] Add quotas to buckets struct - [x] Add CLI to manage bucket quotas - [x] Add admin API to manage bucket quotas - [x] Apply quotas by adding checks on put operations - [x] Proof-read Co-authored-by: Alex Auvolat Reviewed-on: https://git.deuxfleurs.fr/Deuxfleurs/garage/pulls/326 Co-authored-by: Alex Co-committed-by: Alex --- src/db/sled_adapter.rs | 6 ++++++ 1 file changed, 6 insertions(+) (limited to 'src/db/sled_adapter.rs') diff --git a/src/db/sled_adapter.rs b/src/db/sled_adapter.rs index 982f8d82..cf61867d 100644 --- a/src/db/sled_adapter.rs +++ b/src/db/sled_adapter.rs @@ -113,6 +113,12 @@ impl IDb for SledDb { Ok(old_val.map(|x| x.to_vec())) } + fn clear(&self, tree: usize) -> Result<()> { + let tree = self.get_tree(tree)?; + tree.clear()?; + Ok(()) + } + fn iter(&self, tree: usize) -> Result> { let tree = self.get_tree(tree)?; Ok(Box::new(tree.iter().map(|v| { -- cgit v1.2.3