garage/src/table/table.rs

528 lines
14 KiB
Rust
Raw Normal View History

First implementation of K2V (#293) **Specification:** View spec at [this URL](https://git.deuxfleurs.fr/Deuxfleurs/garage/src/branch/k2v/doc/drafts/k2v-spec.md) - [x] Specify the structure of K2V triples - [x] Specify the DVVS format used for causality detection - [x] Specify the K2V index (just a counter of number of values per partition key) - [x] Specify single-item endpoints: ReadItem, InsertItem, DeleteItem - [x] Specify index endpoint: ReadIndex - [x] Specify multi-item endpoints: InsertBatch, ReadBatch, DeleteBatch - [x] Move to JSON objects instead of tuples - [x] Specify endpoints for polling for updates on single values (PollItem) **Implementation:** - [x] Table for K2V items, causal contexts - [x] Indexing mechanism and table for K2V index - [x] Make API handlers a bit more generic - [x] K2V API endpoint - [x] K2V API router - [x] ReadItem - [x] InsertItem - [x] DeleteItem - [x] PollItem - [x] ReadIndex - [x] InsertBatch - [x] ReadBatch - [x] DeleteBatch **Testing:** - [x] Just a simple Python script that does some requests to check visually that things are going right (does not contain parsing of results or assertions on returned values) - [x] Actual tests: - [x] Adapt testing framework - [x] Simple test with InsertItem + ReadItem - [x] Test with several Insert/Read/DeleteItem + ReadIndex - [x] Test all combinations of return formats for ReadItem - [x] Test with ReadBatch, InsertBatch, DeleteBatch - [x] Test with PollItem - [x] Test error codes - [ ] Fix most broken stuff - [x] test PollItem broken randomly - [x] when invalid causality tokens are given, errors should be 4xx not 5xx **Improvements:** - [x] Descending range queries - [x] Specify - [x] Implement - [x] Add test - [x] Batch updates to index counter - [x] Put K2V behind `k2v` feature flag Co-authored-by: Alex Auvolat <alex@adnab.me> Reviewed-on: https://git.deuxfleurs.fr/Deuxfleurs/garage/pulls/293 Co-authored-by: Alex <alex@adnab.me> Co-committed-by: Alex <alex@adnab.me>
2022-05-10 11:16:57 +00:00
use std::borrow::Borrow;
use std::collections::{BTreeMap, BTreeSet, HashMap};
use std::sync::Arc;
2020-04-08 20:00:41 +00:00
2021-10-14 09:50:12 +00:00
use async_trait::async_trait;
use futures::stream::*;
use serde::{Deserialize, Serialize};
use serde_bytes::ByteBuf;
use opentelemetry::{
trace::{FutureExt, TraceContextExt, Tracer},
Context,
};
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 <alex@adnab.me> Reviewed-on: https://git.deuxfleurs.fr/Deuxfleurs/garage/pulls/322 Co-authored-by: Alex <alex@adnab.me> Co-committed-by: Alex <alex@adnab.me>
2022-06-08 08:01:44 +00:00
use garage_db as db;
2022-12-14 15:08:05 +00:00
use garage_util::background::BackgroundRunner;
2020-04-24 10:10:01 +00:00
use garage_util::data::*;
use garage_util::error::Error;
2022-02-22 12:53:59 +00:00
use garage_util::metrics::RecordDuration;
2023-01-03 13:44:47 +00:00
use garage_util::migrate::Migrate;
2020-04-23 17:05:46 +00:00
use garage_rpc::rpc_helper::QuorumSetResultTracker;
2021-10-14 09:50:12 +00:00
use garage_rpc::system::System;
use garage_rpc::*;
2020-04-23 17:05:46 +00:00
2021-05-02 21:13:08 +00:00
use crate::crdt::Crdt;
2021-03-11 15:54:15 +00:00
use crate::data::*;
2021-03-12 20:52:19 +00:00
use crate::gc::*;
2021-03-16 10:43:58 +00:00
use crate::merkle::*;
use crate::queue::InsertQueueWorker;
2021-03-11 15:54:15 +00:00
use crate::replication::*;
use crate::schema::*;
use crate::sync::*;
First implementation of K2V (#293) **Specification:** View spec at [this URL](https://git.deuxfleurs.fr/Deuxfleurs/garage/src/branch/k2v/doc/drafts/k2v-spec.md) - [x] Specify the structure of K2V triples - [x] Specify the DVVS format used for causality detection - [x] Specify the K2V index (just a counter of number of values per partition key) - [x] Specify single-item endpoints: ReadItem, InsertItem, DeleteItem - [x] Specify index endpoint: ReadIndex - [x] Specify multi-item endpoints: InsertBatch, ReadBatch, DeleteBatch - [x] Move to JSON objects instead of tuples - [x] Specify endpoints for polling for updates on single values (PollItem) **Implementation:** - [x] Table for K2V items, causal contexts - [x] Indexing mechanism and table for K2V index - [x] Make API handlers a bit more generic - [x] K2V API endpoint - [x] K2V API router - [x] ReadItem - [x] InsertItem - [x] DeleteItem - [x] PollItem - [x] ReadIndex - [x] InsertBatch - [x] ReadBatch - [x] DeleteBatch **Testing:** - [x] Just a simple Python script that does some requests to check visually that things are going right (does not contain parsing of results or assertions on returned values) - [x] Actual tests: - [x] Adapt testing framework - [x] Simple test with InsertItem + ReadItem - [x] Test with several Insert/Read/DeleteItem + ReadIndex - [x] Test all combinations of return formats for ReadItem - [x] Test with ReadBatch, InsertBatch, DeleteBatch - [x] Test with PollItem - [x] Test error codes - [ ] Fix most broken stuff - [x] test PollItem broken randomly - [x] when invalid causality tokens are given, errors should be 4xx not 5xx **Improvements:** - [x] Descending range queries - [x] Specify - [x] Implement - [x] Add test - [x] Batch updates to index counter - [x] Put K2V behind `k2v` feature flag Co-authored-by: Alex Auvolat <alex@adnab.me> Reviewed-on: https://git.deuxfleurs.fr/Deuxfleurs/garage/pulls/293 Co-authored-by: Alex <alex@adnab.me> Co-committed-by: Alex <alex@adnab.me>
2022-05-10 11:16:57 +00:00
use crate::util::*;
2020-04-08 20:00:41 +00:00
2023-01-03 14:08:37 +00:00
pub struct Table<F: TableSchema, R: TableReplication> {
2021-03-16 10:43:58 +00:00
pub system: Arc<System>,
pub data: Arc<TableData<F, R>>,
pub merkle_updater: Arc<MerkleUpdater<F, R>>,
2021-03-11 15:54:15 +00:00
pub syncer: Arc<TableSyncer<F, R>>,
gc: Arc<TableGc<F, R>>,
2021-10-14 09:50:12 +00:00
endpoint: Arc<Endpoint<TableRpc<F>, Self>>,
2020-04-08 20:00:41 +00:00
}
2020-04-08 21:01:49 +00:00
#[derive(Serialize, Deserialize)]
2021-05-02 21:13:08 +00:00
pub(crate) enum TableRpc<F: TableSchema> {
2020-04-08 21:01:49 +00:00
Ok,
2020-04-08 21:47:34 +00:00
2020-04-09 14:16:27 +00:00
ReadEntry(F::P, F::S),
ReadEntryResponse(Option<ByteBuf>),
2020-04-08 21:47:34 +00:00
// Read range: read all keys in partition P, possibly starting at a certain sort key offset
First implementation of K2V (#293) **Specification:** View spec at [this URL](https://git.deuxfleurs.fr/Deuxfleurs/garage/src/branch/k2v/doc/drafts/k2v-spec.md) - [x] Specify the structure of K2V triples - [x] Specify the DVVS format used for causality detection - [x] Specify the K2V index (just a counter of number of values per partition key) - [x] Specify single-item endpoints: ReadItem, InsertItem, DeleteItem - [x] Specify index endpoint: ReadIndex - [x] Specify multi-item endpoints: InsertBatch, ReadBatch, DeleteBatch - [x] Move to JSON objects instead of tuples - [x] Specify endpoints for polling for updates on single values (PollItem) **Implementation:** - [x] Table for K2V items, causal contexts - [x] Indexing mechanism and table for K2V index - [x] Make API handlers a bit more generic - [x] K2V API endpoint - [x] K2V API router - [x] ReadItem - [x] InsertItem - [x] DeleteItem - [x] PollItem - [x] ReadIndex - [x] InsertBatch - [x] ReadBatch - [x] DeleteBatch **Testing:** - [x] Just a simple Python script that does some requests to check visually that things are going right (does not contain parsing of results or assertions on returned values) - [x] Actual tests: - [x] Adapt testing framework - [x] Simple test with InsertItem + ReadItem - [x] Test with several Insert/Read/DeleteItem + ReadIndex - [x] Test all combinations of return formats for ReadItem - [x] Test with ReadBatch, InsertBatch, DeleteBatch - [x] Test with PollItem - [x] Test error codes - [ ] Fix most broken stuff - [x] test PollItem broken randomly - [x] when invalid causality tokens are given, errors should be 4xx not 5xx **Improvements:** - [x] Descending range queries - [x] Specify - [x] Implement - [x] Add test - [x] Batch updates to index counter - [x] Put K2V behind `k2v` feature flag Co-authored-by: Alex Auvolat <alex@adnab.me> Reviewed-on: https://git.deuxfleurs.fr/Deuxfleurs/garage/pulls/293 Co-authored-by: Alex <alex@adnab.me> Co-committed-by: Alex <alex@adnab.me>
2022-05-10 11:16:57 +00:00
ReadRange {
partition: F::P,
begin_sort_key: Option<F::S>,
filter: Option<F::Filter>,
limit: usize,
enumeration_order: EnumerationOrder,
},
Update(Vec<Arc<ByteBuf>>),
2020-04-08 20:00:41 +00:00
}
impl<F: TableSchema> Rpc for TableRpc<F> {
type Response = Result<TableRpc<F>, Error>;
2021-10-14 09:50:12 +00:00
}
2020-04-18 17:21:34 +00:00
2023-01-03 14:08:37 +00:00
impl<F: TableSchema, R: TableReplication> Table<F, R> {
// =============== PUBLIC INTERFACE FUNCTIONS (new, insert, get, etc) ===============
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 <alex@adnab.me> Reviewed-on: https://git.deuxfleurs.fr/Deuxfleurs/garage/pulls/322 Co-authored-by: Alex <alex@adnab.me> Co-committed-by: Alex <alex@adnab.me>
2022-06-08 08:01:44 +00:00
pub fn new(instance: F, replication: R, system: Arc<System>, db: &db::Db) -> Arc<Self> {
2021-10-14 09:50:12 +00:00
let endpoint = system
.netapp
2021-12-14 11:34:01 +00:00
.endpoint(format!("garage_table/table.rs/Rpc:{}", F::TABLE_NAME));
2020-04-18 17:21:34 +00:00
2021-12-14 11:34:01 +00:00
let data = TableData::new(system.clone(), instance, replication, db);
let merkle_updater = MerkleUpdater::new(data.clone());
2020-04-18 17:21:34 +00:00
let syncer = TableSyncer::new(system.clone(), data.clone(), merkle_updater.clone());
let gc = TableGc::new(system.clone(), data.clone());
2020-04-18 17:21:34 +00:00
system.layout_manager.add_table(F::TABLE_NAME);
2021-03-12 14:07:23 +00:00
let table = Arc::new(Self {
2021-03-16 10:43:58 +00:00
system,
2021-03-12 14:07:23 +00:00
data,
2021-03-16 10:43:58 +00:00
merkle_updater,
gc,
2021-03-12 14:07:23 +00:00
syncer,
2021-10-14 09:50:12 +00:00
endpoint,
2021-03-12 14:07:23 +00:00
});
2021-03-11 15:54:15 +00:00
2021-10-14 09:50:12 +00:00
table.endpoint.set_handler(table.clone());
2020-04-16 12:50:49 +00:00
table
2020-04-08 20:00:41 +00:00
}
pub fn spawn_workers(self: &Arc<Self>, bg: &BackgroundRunner) {
self.merkle_updater.spawn_workers(bg);
self.syncer.spawn_workers(bg);
self.gc.spawn_workers(bg);
bg.spawn_worker(InsertQueueWorker(self.clone()));
}
2020-04-09 14:16:27 +00:00
pub async fn insert(&self, e: &F::E) -> Result<(), Error> {
let tracer = opentelemetry::global::tracer("garage_table");
let span = tracer.start(format!("{} insert", F::TABLE_NAME));
self.insert_internal(e)
.bound_record_duration(&self.data.metrics.put_request_duration)
.with_context(Context::current_with_span(span))
.await?;
self.data.metrics.put_request_counter.add(1);
Ok(())
}
async fn insert_internal(&self, e: &F::E) -> Result<(), Error> {
2020-04-09 14:16:27 +00:00
let hash = e.partition_key().hash();
let who = self.data.replication.write_sets(&hash);
2020-04-08 21:01:49 +00:00
2023-01-03 13:44:47 +00:00
let e_enc = Arc::new(ByteBuf::from(e.encode()?));
2021-05-02 21:13:08 +00:00
let rpc = TableRpc::<F>::Update(vec![e_enc]);
2021-10-14 09:50:12 +00:00
self.system
.rpc_helper()
.try_write_many_sets(
2021-10-14 09:50:12 +00:00
&self.endpoint,
2023-11-15 14:40:44 +00:00
who.as_ref(),
rpc,
2021-10-14 09:50:12 +00:00
RequestStrategy::with_priority(PRIO_NORMAL)
.with_quorum(self.data.replication.write_quorum()),
)
.await?;
2022-02-16 13:23:04 +00:00
2020-04-08 21:01:49 +00:00
Ok(())
}
2020-04-08 20:00:41 +00:00
/// Insert item locally
pub fn queue_insert(&self, tx: &mut db::Transaction, e: &F::E) -> db::TxResult<(), Error> {
self.data.queue_insert(tx, e)
}
pub async fn insert_many<I, IE>(self: &Arc<Self>, entries: I) -> Result<(), Error>
First implementation of K2V (#293) **Specification:** View spec at [this URL](https://git.deuxfleurs.fr/Deuxfleurs/garage/src/branch/k2v/doc/drafts/k2v-spec.md) - [x] Specify the structure of K2V triples - [x] Specify the DVVS format used for causality detection - [x] Specify the K2V index (just a counter of number of values per partition key) - [x] Specify single-item endpoints: ReadItem, InsertItem, DeleteItem - [x] Specify index endpoint: ReadIndex - [x] Specify multi-item endpoints: InsertBatch, ReadBatch, DeleteBatch - [x] Move to JSON objects instead of tuples - [x] Specify endpoints for polling for updates on single values (PollItem) **Implementation:** - [x] Table for K2V items, causal contexts - [x] Indexing mechanism and table for K2V index - [x] Make API handlers a bit more generic - [x] K2V API endpoint - [x] K2V API router - [x] ReadItem - [x] InsertItem - [x] DeleteItem - [x] PollItem - [x] ReadIndex - [x] InsertBatch - [x] ReadBatch - [x] DeleteBatch **Testing:** - [x] Just a simple Python script that does some requests to check visually that things are going right (does not contain parsing of results or assertions on returned values) - [x] Actual tests: - [x] Adapt testing framework - [x] Simple test with InsertItem + ReadItem - [x] Test with several Insert/Read/DeleteItem + ReadIndex - [x] Test all combinations of return formats for ReadItem - [x] Test with ReadBatch, InsertBatch, DeleteBatch - [x] Test with PollItem - [x] Test error codes - [ ] Fix most broken stuff - [x] test PollItem broken randomly - [x] when invalid causality tokens are given, errors should be 4xx not 5xx **Improvements:** - [x] Descending range queries - [x] Specify - [x] Implement - [x] Add test - [x] Batch updates to index counter - [x] Put K2V behind `k2v` feature flag Co-authored-by: Alex Auvolat <alex@adnab.me> Reviewed-on: https://git.deuxfleurs.fr/Deuxfleurs/garage/pulls/293 Co-authored-by: Alex <alex@adnab.me> Co-committed-by: Alex <alex@adnab.me>
2022-05-10 11:16:57 +00:00
where
I: IntoIterator<Item = IE> + Send + Sync,
IE: Borrow<F::E> + Send + Sync,
{
let tracer = opentelemetry::global::tracer("garage_table");
First implementation of K2V (#293) **Specification:** View spec at [this URL](https://git.deuxfleurs.fr/Deuxfleurs/garage/src/branch/k2v/doc/drafts/k2v-spec.md) - [x] Specify the structure of K2V triples - [x] Specify the DVVS format used for causality detection - [x] Specify the K2V index (just a counter of number of values per partition key) - [x] Specify single-item endpoints: ReadItem, InsertItem, DeleteItem - [x] Specify index endpoint: ReadIndex - [x] Specify multi-item endpoints: InsertBatch, ReadBatch, DeleteBatch - [x] Move to JSON objects instead of tuples - [x] Specify endpoints for polling for updates on single values (PollItem) **Implementation:** - [x] Table for K2V items, causal contexts - [x] Indexing mechanism and table for K2V index - [x] Make API handlers a bit more generic - [x] K2V API endpoint - [x] K2V API router - [x] ReadItem - [x] InsertItem - [x] DeleteItem - [x] PollItem - [x] ReadIndex - [x] InsertBatch - [x] ReadBatch - [x] DeleteBatch **Testing:** - [x] Just a simple Python script that does some requests to check visually that things are going right (does not contain parsing of results or assertions on returned values) - [x] Actual tests: - [x] Adapt testing framework - [x] Simple test with InsertItem + ReadItem - [x] Test with several Insert/Read/DeleteItem + ReadIndex - [x] Test all combinations of return formats for ReadItem - [x] Test with ReadBatch, InsertBatch, DeleteBatch - [x] Test with PollItem - [x] Test error codes - [ ] Fix most broken stuff - [x] test PollItem broken randomly - [x] when invalid causality tokens are given, errors should be 4xx not 5xx **Improvements:** - [x] Descending range queries - [x] Specify - [x] Implement - [x] Add test - [x] Batch updates to index counter - [x] Put K2V behind `k2v` feature flag Co-authored-by: Alex Auvolat <alex@adnab.me> Reviewed-on: https://git.deuxfleurs.fr/Deuxfleurs/garage/pulls/293 Co-authored-by: Alex <alex@adnab.me> Co-committed-by: Alex <alex@adnab.me>
2022-05-10 11:16:57 +00:00
let span = tracer.start(format!("{} insert_many", F::TABLE_NAME));
2022-02-22 12:53:59 +00:00
self.insert_many_internal(entries)
.bound_record_duration(&self.data.metrics.put_request_duration)
.with_context(Context::current_with_span(span))
2022-02-22 12:53:59 +00:00
.await?;
2022-02-22 12:53:59 +00:00
self.data.metrics.put_request_counter.add(1);
2022-02-22 12:53:59 +00:00
Ok(())
}
2022-02-16 13:23:04 +00:00
async fn insert_many_internal<I, IE>(self: &Arc<Self>, entries: I) -> Result<(), Error>
First implementation of K2V (#293) **Specification:** View spec at [this URL](https://git.deuxfleurs.fr/Deuxfleurs/garage/src/branch/k2v/doc/drafts/k2v-spec.md) - [x] Specify the structure of K2V triples - [x] Specify the DVVS format used for causality detection - [x] Specify the K2V index (just a counter of number of values per partition key) - [x] Specify single-item endpoints: ReadItem, InsertItem, DeleteItem - [x] Specify index endpoint: ReadIndex - [x] Specify multi-item endpoints: InsertBatch, ReadBatch, DeleteBatch - [x] Move to JSON objects instead of tuples - [x] Specify endpoints for polling for updates on single values (PollItem) **Implementation:** - [x] Table for K2V items, causal contexts - [x] Indexing mechanism and table for K2V index - [x] Make API handlers a bit more generic - [x] K2V API endpoint - [x] K2V API router - [x] ReadItem - [x] InsertItem - [x] DeleteItem - [x] PollItem - [x] ReadIndex - [x] InsertBatch - [x] ReadBatch - [x] DeleteBatch **Testing:** - [x] Just a simple Python script that does some requests to check visually that things are going right (does not contain parsing of results or assertions on returned values) - [x] Actual tests: - [x] Adapt testing framework - [x] Simple test with InsertItem + ReadItem - [x] Test with several Insert/Read/DeleteItem + ReadIndex - [x] Test all combinations of return formats for ReadItem - [x] Test with ReadBatch, InsertBatch, DeleteBatch - [x] Test with PollItem - [x] Test error codes - [ ] Fix most broken stuff - [x] test PollItem broken randomly - [x] when invalid causality tokens are given, errors should be 4xx not 5xx **Improvements:** - [x] Descending range queries - [x] Specify - [x] Implement - [x] Add test - [x] Batch updates to index counter - [x] Put K2V behind `k2v` feature flag Co-authored-by: Alex Auvolat <alex@adnab.me> Reviewed-on: https://git.deuxfleurs.fr/Deuxfleurs/garage/pulls/293 Co-authored-by: Alex <alex@adnab.me> Co-committed-by: Alex <alex@adnab.me>
2022-05-10 11:16:57 +00:00
where
I: IntoIterator<Item = IE> + Send + Sync,
IE: Borrow<F::E> + Send + Sync,
{
// The different items will have to be stored on possibly different nodes.
// We will here batch all items into a single request for each concerned
// node, with all of the entries it must store within that request.
// Each entry has to be saved to a specific list of "write sets", i.e. a set
// of node within wich a quorum must be achieved. In normal operation, there
// is a single write set which corresponds to the quorum in the current
// cluster layout, but when the layout is updated, multiple write sets might
// have to be handled at once. Here, since we are sending many entries, we
// will have to handle many write sets in all cases. The algorihtm is thus
// to send one request to each node with all the items it must save,
// and keep track of the OK responses within each write set: if for all sets
// a quorum of nodes has answered OK, then the insert has succeeded and
// consistency properties (read-after-write) are preserved.
let quorum = self.data.replication.write_quorum();
// Serialize all entries and compute the write sets for each of them.
// In the case of sharded table replication, this also takes an "ack lock"
// to the layout manager to avoid ack'ing newer versions which are not
// taken into account by writes in progress (the ack can happen later, once
// all writes that didn't take the new layout into account are finished).
// These locks are released when entries_vec is dropped, i.e. when this
// function returns.
let mut entries_vec = Vec::new();
First implementation of K2V (#293) **Specification:** View spec at [this URL](https://git.deuxfleurs.fr/Deuxfleurs/garage/src/branch/k2v/doc/drafts/k2v-spec.md) - [x] Specify the structure of K2V triples - [x] Specify the DVVS format used for causality detection - [x] Specify the K2V index (just a counter of number of values per partition key) - [x] Specify single-item endpoints: ReadItem, InsertItem, DeleteItem - [x] Specify index endpoint: ReadIndex - [x] Specify multi-item endpoints: InsertBatch, ReadBatch, DeleteBatch - [x] Move to JSON objects instead of tuples - [x] Specify endpoints for polling for updates on single values (PollItem) **Implementation:** - [x] Table for K2V items, causal contexts - [x] Indexing mechanism and table for K2V index - [x] Make API handlers a bit more generic - [x] K2V API endpoint - [x] K2V API router - [x] ReadItem - [x] InsertItem - [x] DeleteItem - [x] PollItem - [x] ReadIndex - [x] InsertBatch - [x] ReadBatch - [x] DeleteBatch **Testing:** - [x] Just a simple Python script that does some requests to check visually that things are going right (does not contain parsing of results or assertions on returned values) - [x] Actual tests: - [x] Adapt testing framework - [x] Simple test with InsertItem + ReadItem - [x] Test with several Insert/Read/DeleteItem + ReadIndex - [x] Test all combinations of return formats for ReadItem - [x] Test with ReadBatch, InsertBatch, DeleteBatch - [x] Test with PollItem - [x] Test error codes - [ ] Fix most broken stuff - [x] test PollItem broken randomly - [x] when invalid causality tokens are given, errors should be 4xx not 5xx **Improvements:** - [x] Descending range queries - [x] Specify - [x] Implement - [x] Add test - [x] Batch updates to index counter - [x] Put K2V behind `k2v` feature flag Co-authored-by: Alex Auvolat <alex@adnab.me> Reviewed-on: https://git.deuxfleurs.fr/Deuxfleurs/garage/pulls/293 Co-authored-by: Alex <alex@adnab.me> Co-committed-by: Alex <alex@adnab.me>
2022-05-10 11:16:57 +00:00
for entry in entries.into_iter() {
let entry = entry.borrow();
let hash = entry.partition_key().hash();
let mut write_sets = self.data.replication.write_sets(&hash);
for set in write_sets.as_mut().iter_mut() {
set.sort();
}
2023-01-03 13:44:47 +00:00
let e_enc = Arc::new(ByteBuf::from(entry.encode()?));
entries_vec.push((write_sets, e_enc));
}
// Compute a deduplicated list of all of the write sets,
// and compute an index from each node to the position of the sets in which
// it takes part, to optimize the detection of a quorum.
let mut write_sets = entries_vec
.iter()
.map(|(wss, _)| wss.as_ref().iter().map(|ws| ws.as_slice()))
.flatten()
.collect::<Vec<&[Uuid]>>();
write_sets.sort();
write_sets.dedup();
let mut result_tracker = QuorumSetResultTracker::new(&write_sets, quorum);
// Build a map of all nodes to the entries that must be sent to that node.
let mut call_list: HashMap<Uuid, Vec<_>> = HashMap::new();
for (write_sets, entry_enc) in entries_vec.iter() {
for write_set in write_sets.as_ref().iter() {
for node in write_set.iter() {
call_list.entry(*node).or_default().push(entry_enc.clone())
}
}
}
// Build futures to actually perform each of the corresponding RPC calls
let call_futures = call_list.into_iter().map(|(node, entries)| {
let this = self.clone();
let tracer = opentelemetry::global::tracer("garage");
let span = tracer.start(format!("RPC to {:?}", node));
let fut = async move {
let rpc = TableRpc::<F>::Update(entries);
let resp = this
.system
.rpc_helper()
.call(
&this.endpoint,
node,
rpc,
RequestStrategy::with_priority(PRIO_NORMAL).with_quorum(quorum),
)
.await;
(node, resp)
};
fut.with_context(Context::current_with_span(span))
});
// Run all requests in parallel thanks to FuturesUnordered, and collect results.
let mut resps = call_futures.collect::<FuturesUnordered<_>>();
while let Some((node, resp)) = resps.next().await {
result_tracker.register_result(node, resp.map(|_| ()));
if result_tracker.all_quorums_ok() {
// Success
// Continue all other requests in background
tokio::spawn(async move {
resps.collect::<Vec<(Uuid, Result<_, _>)>>().await;
});
return Ok(());
}
if result_tracker.too_many_failures() {
// Too many errors in this set, we know we won't get a quorum
break;
}
}
// Failure, could not get quorum within at least one set
Err(result_tracker.quorum_error())
}
2020-04-16 12:50:49 +00:00
pub async fn get(
self: &Arc<Self>,
partition_key: &F::P,
sort_key: &F::S,
) -> Result<Option<F::E>, Error> {
let tracer = opentelemetry::global::tracer("garage_table");
let span = tracer.start(format!("{} get", F::TABLE_NAME));
2022-02-22 13:52:41 +00:00
let res = self
.get_internal(partition_key, sort_key)
2022-02-22 12:53:59 +00:00
.bound_record_duration(&self.data.metrics.get_request_duration)
.with_context(Context::current_with_span(span))
2022-02-22 12:53:59 +00:00
.await?;
2022-02-22 12:53:59 +00:00
self.data.metrics.get_request_counter.add(1);
2022-02-22 12:53:59 +00:00
Ok(res)
}
2022-02-16 13:23:04 +00:00
2022-02-22 12:53:59 +00:00
async fn get_internal(
self: &Arc<Self>,
partition_key: &F::P,
sort_key: &F::S,
) -> Result<Option<F::E>, Error> {
2020-04-09 14:16:27 +00:00
let hash = partition_key.hash();
2021-03-16 10:43:58 +00:00
let who = self.data.replication.read_nodes(&hash);
2020-04-08 20:00:41 +00:00
2021-05-02 21:13:08 +00:00
let rpc = TableRpc::<F>::ReadEntry(partition_key.clone(), sort_key.clone());
let resps = self
2021-10-14 09:50:12 +00:00
.system
.rpc_helper()
.try_call_many(
2021-10-14 09:50:12 +00:00
&self.endpoint,
&who,
rpc,
2021-10-14 09:50:12 +00:00
RequestStrategy::with_priority(PRIO_NORMAL)
.with_quorum(self.data.replication.read_quorum()),
)
2020-04-08 21:01:49 +00:00
.await?;
2020-04-08 21:47:34 +00:00
let mut ret = None;
2020-04-09 18:58:39 +00:00
let mut not_all_same = false;
2020-04-08 21:01:49 +00:00
for resp in resps {
2021-05-02 21:13:08 +00:00
if let TableRpc::ReadEntryResponse(value) = resp {
if let Some(v_bytes) = value {
2021-03-11 15:54:15 +00:00
let v = self.data.decode_entry(v_bytes.as_slice())?;
2020-04-08 21:47:34 +00:00
ret = match ret {
None => Some(v),
Some(mut x) => {
2020-04-09 21:45:07 +00:00
if x != v {
2020-04-09 18:58:39 +00:00
not_all_same = true;
2020-04-09 21:45:07 +00:00
x.merge(&v);
2020-04-09 18:58:39 +00:00
}
2020-04-08 21:47:34 +00:00
Some(x)
}
}
2020-04-08 21:01:49 +00:00
}
2020-04-08 21:47:34 +00:00
} else {
2021-04-23 19:42:52 +00:00
return Err(Error::Message("Invalid return value to read".to_string()));
2020-04-08 21:01:49 +00:00
}
}
2020-04-09 18:58:39 +00:00
if let Some(ret_entry) = &ret {
if not_all_same {
let self2 = self.clone();
let ent2 = ret_entry.clone();
2022-12-14 15:08:05 +00:00
tokio::spawn(async move {
if let Err(e) = self2.repair_on_read(&who[..], ent2).await {
warn!("Error doing repair on read: {}", e);
}
});
2020-04-09 18:58:39 +00:00
}
}
2022-02-16 13:23:04 +00:00
2020-04-08 21:47:34 +00:00
Ok(ret)
2020-04-08 21:01:49 +00:00
}
pub async fn get_range(
self: &Arc<Self>,
partition_key: &F::P,
begin_sort_key: Option<F::S>,
filter: Option<F::Filter>,
limit: usize,
First implementation of K2V (#293) **Specification:** View spec at [this URL](https://git.deuxfleurs.fr/Deuxfleurs/garage/src/branch/k2v/doc/drafts/k2v-spec.md) - [x] Specify the structure of K2V triples - [x] Specify the DVVS format used for causality detection - [x] Specify the K2V index (just a counter of number of values per partition key) - [x] Specify single-item endpoints: ReadItem, InsertItem, DeleteItem - [x] Specify index endpoint: ReadIndex - [x] Specify multi-item endpoints: InsertBatch, ReadBatch, DeleteBatch - [x] Move to JSON objects instead of tuples - [x] Specify endpoints for polling for updates on single values (PollItem) **Implementation:** - [x] Table for K2V items, causal contexts - [x] Indexing mechanism and table for K2V index - [x] Make API handlers a bit more generic - [x] K2V API endpoint - [x] K2V API router - [x] ReadItem - [x] InsertItem - [x] DeleteItem - [x] PollItem - [x] ReadIndex - [x] InsertBatch - [x] ReadBatch - [x] DeleteBatch **Testing:** - [x] Just a simple Python script that does some requests to check visually that things are going right (does not contain parsing of results or assertions on returned values) - [x] Actual tests: - [x] Adapt testing framework - [x] Simple test with InsertItem + ReadItem - [x] Test with several Insert/Read/DeleteItem + ReadIndex - [x] Test all combinations of return formats for ReadItem - [x] Test with ReadBatch, InsertBatch, DeleteBatch - [x] Test with PollItem - [x] Test error codes - [ ] Fix most broken stuff - [x] test PollItem broken randomly - [x] when invalid causality tokens are given, errors should be 4xx not 5xx **Improvements:** - [x] Descending range queries - [x] Specify - [x] Implement - [x] Add test - [x] Batch updates to index counter - [x] Put K2V behind `k2v` feature flag Co-authored-by: Alex Auvolat <alex@adnab.me> Reviewed-on: https://git.deuxfleurs.fr/Deuxfleurs/garage/pulls/293 Co-authored-by: Alex <alex@adnab.me> Co-committed-by: Alex <alex@adnab.me>
2022-05-10 11:16:57 +00:00
enumeration_order: EnumerationOrder,
) -> Result<Vec<F::E>, Error> {
let tracer = opentelemetry::global::tracer("garage_table");
let span = tracer.start(format!("{} get_range", F::TABLE_NAME));
2022-02-22 13:52:41 +00:00
let res = self
First implementation of K2V (#293) **Specification:** View spec at [this URL](https://git.deuxfleurs.fr/Deuxfleurs/garage/src/branch/k2v/doc/drafts/k2v-spec.md) - [x] Specify the structure of K2V triples - [x] Specify the DVVS format used for causality detection - [x] Specify the K2V index (just a counter of number of values per partition key) - [x] Specify single-item endpoints: ReadItem, InsertItem, DeleteItem - [x] Specify index endpoint: ReadIndex - [x] Specify multi-item endpoints: InsertBatch, ReadBatch, DeleteBatch - [x] Move to JSON objects instead of tuples - [x] Specify endpoints for polling for updates on single values (PollItem) **Implementation:** - [x] Table for K2V items, causal contexts - [x] Indexing mechanism and table for K2V index - [x] Make API handlers a bit more generic - [x] K2V API endpoint - [x] K2V API router - [x] ReadItem - [x] InsertItem - [x] DeleteItem - [x] PollItem - [x] ReadIndex - [x] InsertBatch - [x] ReadBatch - [x] DeleteBatch **Testing:** - [x] Just a simple Python script that does some requests to check visually that things are going right (does not contain parsing of results or assertions on returned values) - [x] Actual tests: - [x] Adapt testing framework - [x] Simple test with InsertItem + ReadItem - [x] Test with several Insert/Read/DeleteItem + ReadIndex - [x] Test all combinations of return formats for ReadItem - [x] Test with ReadBatch, InsertBatch, DeleteBatch - [x] Test with PollItem - [x] Test error codes - [ ] Fix most broken stuff - [x] test PollItem broken randomly - [x] when invalid causality tokens are given, errors should be 4xx not 5xx **Improvements:** - [x] Descending range queries - [x] Specify - [x] Implement - [x] Add test - [x] Batch updates to index counter - [x] Put K2V behind `k2v` feature flag Co-authored-by: Alex Auvolat <alex@adnab.me> Reviewed-on: https://git.deuxfleurs.fr/Deuxfleurs/garage/pulls/293 Co-authored-by: Alex <alex@adnab.me> Co-committed-by: Alex <alex@adnab.me>
2022-05-10 11:16:57 +00:00
.get_range_internal(
partition_key,
begin_sort_key,
filter,
limit,
enumeration_order,
)
2022-02-22 12:53:59 +00:00
.bound_record_duration(&self.data.metrics.get_request_duration)
.with_context(Context::current_with_span(span))
2022-02-22 12:53:59 +00:00
.await?;
2022-02-22 12:53:59 +00:00
self.data.metrics.get_request_counter.add(1);
2022-02-22 12:53:59 +00:00
Ok(res)
}
2022-02-16 13:23:04 +00:00
2022-02-22 12:53:59 +00:00
async fn get_range_internal(
self: &Arc<Self>,
partition_key: &F::P,
begin_sort_key: Option<F::S>,
filter: Option<F::Filter>,
limit: usize,
First implementation of K2V (#293) **Specification:** View spec at [this URL](https://git.deuxfleurs.fr/Deuxfleurs/garage/src/branch/k2v/doc/drafts/k2v-spec.md) - [x] Specify the structure of K2V triples - [x] Specify the DVVS format used for causality detection - [x] Specify the K2V index (just a counter of number of values per partition key) - [x] Specify single-item endpoints: ReadItem, InsertItem, DeleteItem - [x] Specify index endpoint: ReadIndex - [x] Specify multi-item endpoints: InsertBatch, ReadBatch, DeleteBatch - [x] Move to JSON objects instead of tuples - [x] Specify endpoints for polling for updates on single values (PollItem) **Implementation:** - [x] Table for K2V items, causal contexts - [x] Indexing mechanism and table for K2V index - [x] Make API handlers a bit more generic - [x] K2V API endpoint - [x] K2V API router - [x] ReadItem - [x] InsertItem - [x] DeleteItem - [x] PollItem - [x] ReadIndex - [x] InsertBatch - [x] ReadBatch - [x] DeleteBatch **Testing:** - [x] Just a simple Python script that does some requests to check visually that things are going right (does not contain parsing of results or assertions on returned values) - [x] Actual tests: - [x] Adapt testing framework - [x] Simple test with InsertItem + ReadItem - [x] Test with several Insert/Read/DeleteItem + ReadIndex - [x] Test all combinations of return formats for ReadItem - [x] Test with ReadBatch, InsertBatch, DeleteBatch - [x] Test with PollItem - [x] Test error codes - [ ] Fix most broken stuff - [x] test PollItem broken randomly - [x] when invalid causality tokens are given, errors should be 4xx not 5xx **Improvements:** - [x] Descending range queries - [x] Specify - [x] Implement - [x] Add test - [x] Batch updates to index counter - [x] Put K2V behind `k2v` feature flag Co-authored-by: Alex Auvolat <alex@adnab.me> Reviewed-on: https://git.deuxfleurs.fr/Deuxfleurs/garage/pulls/293 Co-authored-by: Alex <alex@adnab.me> Co-committed-by: Alex <alex@adnab.me>
2022-05-10 11:16:57 +00:00
enumeration_order: EnumerationOrder,
2022-02-22 12:53:59 +00:00
) -> Result<Vec<F::E>, Error> {
let hash = partition_key.hash();
2021-03-16 10:43:58 +00:00
let who = self.data.replication.read_nodes(&hash);
First implementation of K2V (#293) **Specification:** View spec at [this URL](https://git.deuxfleurs.fr/Deuxfleurs/garage/src/branch/k2v/doc/drafts/k2v-spec.md) - [x] Specify the structure of K2V triples - [x] Specify the DVVS format used for causality detection - [x] Specify the K2V index (just a counter of number of values per partition key) - [x] Specify single-item endpoints: ReadItem, InsertItem, DeleteItem - [x] Specify index endpoint: ReadIndex - [x] Specify multi-item endpoints: InsertBatch, ReadBatch, DeleteBatch - [x] Move to JSON objects instead of tuples - [x] Specify endpoints for polling for updates on single values (PollItem) **Implementation:** - [x] Table for K2V items, causal contexts - [x] Indexing mechanism and table for K2V index - [x] Make API handlers a bit more generic - [x] K2V API endpoint - [x] K2V API router - [x] ReadItem - [x] InsertItem - [x] DeleteItem - [x] PollItem - [x] ReadIndex - [x] InsertBatch - [x] ReadBatch - [x] DeleteBatch **Testing:** - [x] Just a simple Python script that does some requests to check visually that things are going right (does not contain parsing of results or assertions on returned values) - [x] Actual tests: - [x] Adapt testing framework - [x] Simple test with InsertItem + ReadItem - [x] Test with several Insert/Read/DeleteItem + ReadIndex - [x] Test all combinations of return formats for ReadItem - [x] Test with ReadBatch, InsertBatch, DeleteBatch - [x] Test with PollItem - [x] Test error codes - [ ] Fix most broken stuff - [x] test PollItem broken randomly - [x] when invalid causality tokens are given, errors should be 4xx not 5xx **Improvements:** - [x] Descending range queries - [x] Specify - [x] Implement - [x] Add test - [x] Batch updates to index counter - [x] Put K2V behind `k2v` feature flag Co-authored-by: Alex Auvolat <alex@adnab.me> Reviewed-on: https://git.deuxfleurs.fr/Deuxfleurs/garage/pulls/293 Co-authored-by: Alex <alex@adnab.me> Co-committed-by: Alex <alex@adnab.me>
2022-05-10 11:16:57 +00:00
let rpc = TableRpc::<F>::ReadRange {
partition: partition_key.clone(),
begin_sort_key,
filter,
limit,
enumeration_order,
};
let resps = self
2021-10-14 09:50:12 +00:00
.system
.rpc_helper()
.try_call_many(
2021-10-14 09:50:12 +00:00
&self.endpoint,
&who,
rpc,
2021-10-14 09:50:12 +00:00
RequestStrategy::with_priority(PRIO_NORMAL)
.with_quorum(self.data.replication.read_quorum()),
)
.await?;
First implementation of K2V (#293) **Specification:** View spec at [this URL](https://git.deuxfleurs.fr/Deuxfleurs/garage/src/branch/k2v/doc/drafts/k2v-spec.md) - [x] Specify the structure of K2V triples - [x] Specify the DVVS format used for causality detection - [x] Specify the K2V index (just a counter of number of values per partition key) - [x] Specify single-item endpoints: ReadItem, InsertItem, DeleteItem - [x] Specify index endpoint: ReadIndex - [x] Specify multi-item endpoints: InsertBatch, ReadBatch, DeleteBatch - [x] Move to JSON objects instead of tuples - [x] Specify endpoints for polling for updates on single values (PollItem) **Implementation:** - [x] Table for K2V items, causal contexts - [x] Indexing mechanism and table for K2V index - [x] Make API handlers a bit more generic - [x] K2V API endpoint - [x] K2V API router - [x] ReadItem - [x] InsertItem - [x] DeleteItem - [x] PollItem - [x] ReadIndex - [x] InsertBatch - [x] ReadBatch - [x] DeleteBatch **Testing:** - [x] Just a simple Python script that does some requests to check visually that things are going right (does not contain parsing of results or assertions on returned values) - [x] Actual tests: - [x] Adapt testing framework - [x] Simple test with InsertItem + ReadItem - [x] Test with several Insert/Read/DeleteItem + ReadIndex - [x] Test all combinations of return formats for ReadItem - [x] Test with ReadBatch, InsertBatch, DeleteBatch - [x] Test with PollItem - [x] Test error codes - [ ] Fix most broken stuff - [x] test PollItem broken randomly - [x] when invalid causality tokens are given, errors should be 4xx not 5xx **Improvements:** - [x] Descending range queries - [x] Specify - [x] Implement - [x] Add test - [x] Batch updates to index counter - [x] Put K2V behind `k2v` feature flag Co-authored-by: Alex Auvolat <alex@adnab.me> Reviewed-on: https://git.deuxfleurs.fr/Deuxfleurs/garage/pulls/293 Co-authored-by: Alex <alex@adnab.me> Co-committed-by: Alex <alex@adnab.me>
2022-05-10 11:16:57 +00:00
let mut ret: BTreeMap<Vec<u8>, F::E> = BTreeMap::new();
let mut to_repair = BTreeSet::new();
for resp in resps {
2021-05-02 21:13:08 +00:00
if let TableRpc::Update(entries) = resp {
for entry_bytes in entries.iter() {
2021-03-11 15:54:15 +00:00
let entry = self.data.decode_entry(entry_bytes.as_slice())?;
let entry_key = self.data.tree_key(entry.partition_key(), entry.sort_key());
First implementation of K2V (#293) **Specification:** View spec at [this URL](https://git.deuxfleurs.fr/Deuxfleurs/garage/src/branch/k2v/doc/drafts/k2v-spec.md) - [x] Specify the structure of K2V triples - [x] Specify the DVVS format used for causality detection - [x] Specify the K2V index (just a counter of number of values per partition key) - [x] Specify single-item endpoints: ReadItem, InsertItem, DeleteItem - [x] Specify index endpoint: ReadIndex - [x] Specify multi-item endpoints: InsertBatch, ReadBatch, DeleteBatch - [x] Move to JSON objects instead of tuples - [x] Specify endpoints for polling for updates on single values (PollItem) **Implementation:** - [x] Table for K2V items, causal contexts - [x] Indexing mechanism and table for K2V index - [x] Make API handlers a bit more generic - [x] K2V API endpoint - [x] K2V API router - [x] ReadItem - [x] InsertItem - [x] DeleteItem - [x] PollItem - [x] ReadIndex - [x] InsertBatch - [x] ReadBatch - [x] DeleteBatch **Testing:** - [x] Just a simple Python script that does some requests to check visually that things are going right (does not contain parsing of results or assertions on returned values) - [x] Actual tests: - [x] Adapt testing framework - [x] Simple test with InsertItem + ReadItem - [x] Test with several Insert/Read/DeleteItem + ReadIndex - [x] Test all combinations of return formats for ReadItem - [x] Test with ReadBatch, InsertBatch, DeleteBatch - [x] Test with PollItem - [x] Test error codes - [ ] Fix most broken stuff - [x] test PollItem broken randomly - [x] when invalid causality tokens are given, errors should be 4xx not 5xx **Improvements:** - [x] Descending range queries - [x] Specify - [x] Implement - [x] Add test - [x] Batch updates to index counter - [x] Put K2V behind `k2v` feature flag Co-authored-by: Alex Auvolat <alex@adnab.me> Reviewed-on: https://git.deuxfleurs.fr/Deuxfleurs/garage/pulls/293 Co-authored-by: Alex <alex@adnab.me> Co-committed-by: Alex <alex@adnab.me>
2022-05-10 11:16:57 +00:00
match ret.get_mut(&entry_key) {
Some(e) => {
if *e != entry {
e.merge(&entry);
to_repair.insert(entry_key.clone());
}
}
First implementation of K2V (#293) **Specification:** View spec at [this URL](https://git.deuxfleurs.fr/Deuxfleurs/garage/src/branch/k2v/doc/drafts/k2v-spec.md) - [x] Specify the structure of K2V triples - [x] Specify the DVVS format used for causality detection - [x] Specify the K2V index (just a counter of number of values per partition key) - [x] Specify single-item endpoints: ReadItem, InsertItem, DeleteItem - [x] Specify index endpoint: ReadIndex - [x] Specify multi-item endpoints: InsertBatch, ReadBatch, DeleteBatch - [x] Move to JSON objects instead of tuples - [x] Specify endpoints for polling for updates on single values (PollItem) **Implementation:** - [x] Table for K2V items, causal contexts - [x] Indexing mechanism and table for K2V index - [x] Make API handlers a bit more generic - [x] K2V API endpoint - [x] K2V API router - [x] ReadItem - [x] InsertItem - [x] DeleteItem - [x] PollItem - [x] ReadIndex - [x] InsertBatch - [x] ReadBatch - [x] DeleteBatch **Testing:** - [x] Just a simple Python script that does some requests to check visually that things are going right (does not contain parsing of results or assertions on returned values) - [x] Actual tests: - [x] Adapt testing framework - [x] Simple test with InsertItem + ReadItem - [x] Test with several Insert/Read/DeleteItem + ReadIndex - [x] Test all combinations of return formats for ReadItem - [x] Test with ReadBatch, InsertBatch, DeleteBatch - [x] Test with PollItem - [x] Test error codes - [ ] Fix most broken stuff - [x] test PollItem broken randomly - [x] when invalid causality tokens are given, errors should be 4xx not 5xx **Improvements:** - [x] Descending range queries - [x] Specify - [x] Implement - [x] Add test - [x] Batch updates to index counter - [x] Put K2V behind `k2v` feature flag Co-authored-by: Alex Auvolat <alex@adnab.me> Reviewed-on: https://git.deuxfleurs.fr/Deuxfleurs/garage/pulls/293 Co-authored-by: Alex <alex@adnab.me> Co-committed-by: Alex <alex@adnab.me>
2022-05-10 11:16:57 +00:00
None => {
ret.insert(entry_key, entry);
}
}
}
First implementation of K2V (#293) **Specification:** View spec at [this URL](https://git.deuxfleurs.fr/Deuxfleurs/garage/src/branch/k2v/doc/drafts/k2v-spec.md) - [x] Specify the structure of K2V triples - [x] Specify the DVVS format used for causality detection - [x] Specify the K2V index (just a counter of number of values per partition key) - [x] Specify single-item endpoints: ReadItem, InsertItem, DeleteItem - [x] Specify index endpoint: ReadIndex - [x] Specify multi-item endpoints: InsertBatch, ReadBatch, DeleteBatch - [x] Move to JSON objects instead of tuples - [x] Specify endpoints for polling for updates on single values (PollItem) **Implementation:** - [x] Table for K2V items, causal contexts - [x] Indexing mechanism and table for K2V index - [x] Make API handlers a bit more generic - [x] K2V API endpoint - [x] K2V API router - [x] ReadItem - [x] InsertItem - [x] DeleteItem - [x] PollItem - [x] ReadIndex - [x] InsertBatch - [x] ReadBatch - [x] DeleteBatch **Testing:** - [x] Just a simple Python script that does some requests to check visually that things are going right (does not contain parsing of results or assertions on returned values) - [x] Actual tests: - [x] Adapt testing framework - [x] Simple test with InsertItem + ReadItem - [x] Test with several Insert/Read/DeleteItem + ReadIndex - [x] Test all combinations of return formats for ReadItem - [x] Test with ReadBatch, InsertBatch, DeleteBatch - [x] Test with PollItem - [x] Test error codes - [ ] Fix most broken stuff - [x] test PollItem broken randomly - [x] when invalid causality tokens are given, errors should be 4xx not 5xx **Improvements:** - [x] Descending range queries - [x] Specify - [x] Implement - [x] Add test - [x] Batch updates to index counter - [x] Put K2V behind `k2v` feature flag Co-authored-by: Alex Auvolat <alex@adnab.me> Reviewed-on: https://git.deuxfleurs.fr/Deuxfleurs/garage/pulls/293 Co-authored-by: Alex <alex@adnab.me> Co-committed-by: Alex <alex@adnab.me>
2022-05-10 11:16:57 +00:00
} else {
return Err(Error::unexpected_rpc_message(resp));
}
}
First implementation of K2V (#293) **Specification:** View spec at [this URL](https://git.deuxfleurs.fr/Deuxfleurs/garage/src/branch/k2v/doc/drafts/k2v-spec.md) - [x] Specify the structure of K2V triples - [x] Specify the DVVS format used for causality detection - [x] Specify the K2V index (just a counter of number of values per partition key) - [x] Specify single-item endpoints: ReadItem, InsertItem, DeleteItem - [x] Specify index endpoint: ReadIndex - [x] Specify multi-item endpoints: InsertBatch, ReadBatch, DeleteBatch - [x] Move to JSON objects instead of tuples - [x] Specify endpoints for polling for updates on single values (PollItem) **Implementation:** - [x] Table for K2V items, causal contexts - [x] Indexing mechanism and table for K2V index - [x] Make API handlers a bit more generic - [x] K2V API endpoint - [x] K2V API router - [x] ReadItem - [x] InsertItem - [x] DeleteItem - [x] PollItem - [x] ReadIndex - [x] InsertBatch - [x] ReadBatch - [x] DeleteBatch **Testing:** - [x] Just a simple Python script that does some requests to check visually that things are going right (does not contain parsing of results or assertions on returned values) - [x] Actual tests: - [x] Adapt testing framework - [x] Simple test with InsertItem + ReadItem - [x] Test with several Insert/Read/DeleteItem + ReadIndex - [x] Test all combinations of return formats for ReadItem - [x] Test with ReadBatch, InsertBatch, DeleteBatch - [x] Test with PollItem - [x] Test error codes - [ ] Fix most broken stuff - [x] test PollItem broken randomly - [x] when invalid causality tokens are given, errors should be 4xx not 5xx **Improvements:** - [x] Descending range queries - [x] Specify - [x] Implement - [x] Add test - [x] Batch updates to index counter - [x] Put K2V behind `k2v` feature flag Co-authored-by: Alex Auvolat <alex@adnab.me> Reviewed-on: https://git.deuxfleurs.fr/Deuxfleurs/garage/pulls/293 Co-authored-by: Alex <alex@adnab.me> Co-committed-by: Alex <alex@adnab.me>
2022-05-10 11:16:57 +00:00
if !to_repair.is_empty() {
let self2 = self.clone();
First implementation of K2V (#293) **Specification:** View spec at [this URL](https://git.deuxfleurs.fr/Deuxfleurs/garage/src/branch/k2v/doc/drafts/k2v-spec.md) - [x] Specify the structure of K2V triples - [x] Specify the DVVS format used for causality detection - [x] Specify the K2V index (just a counter of number of values per partition key) - [x] Specify single-item endpoints: ReadItem, InsertItem, DeleteItem - [x] Specify index endpoint: ReadIndex - [x] Specify multi-item endpoints: InsertBatch, ReadBatch, DeleteBatch - [x] Move to JSON objects instead of tuples - [x] Specify endpoints for polling for updates on single values (PollItem) **Implementation:** - [x] Table for K2V items, causal contexts - [x] Indexing mechanism and table for K2V index - [x] Make API handlers a bit more generic - [x] K2V API endpoint - [x] K2V API router - [x] ReadItem - [x] InsertItem - [x] DeleteItem - [x] PollItem - [x] ReadIndex - [x] InsertBatch - [x] ReadBatch - [x] DeleteBatch **Testing:** - [x] Just a simple Python script that does some requests to check visually that things are going right (does not contain parsing of results or assertions on returned values) - [x] Actual tests: - [x] Adapt testing framework - [x] Simple test with InsertItem + ReadItem - [x] Test with several Insert/Read/DeleteItem + ReadIndex - [x] Test all combinations of return formats for ReadItem - [x] Test with ReadBatch, InsertBatch, DeleteBatch - [x] Test with PollItem - [x] Test error codes - [ ] Fix most broken stuff - [x] test PollItem broken randomly - [x] when invalid causality tokens are given, errors should be 4xx not 5xx **Improvements:** - [x] Descending range queries - [x] Specify - [x] Implement - [x] Add test - [x] Batch updates to index counter - [x] Put K2V behind `k2v` feature flag Co-authored-by: Alex Auvolat <alex@adnab.me> Reviewed-on: https://git.deuxfleurs.fr/Deuxfleurs/garage/pulls/293 Co-authored-by: Alex <alex@adnab.me> Co-committed-by: Alex <alex@adnab.me>
2022-05-10 11:16:57 +00:00
let to_repair = to_repair
.into_iter()
.map(|k| ret.get(&k).unwrap().clone())
.collect::<Vec<_>>();
2022-12-14 15:08:05 +00:00
tokio::spawn(async move {
First implementation of K2V (#293) **Specification:** View spec at [this URL](https://git.deuxfleurs.fr/Deuxfleurs/garage/src/branch/k2v/doc/drafts/k2v-spec.md) - [x] Specify the structure of K2V triples - [x] Specify the DVVS format used for causality detection - [x] Specify the K2V index (just a counter of number of values per partition key) - [x] Specify single-item endpoints: ReadItem, InsertItem, DeleteItem - [x] Specify index endpoint: ReadIndex - [x] Specify multi-item endpoints: InsertBatch, ReadBatch, DeleteBatch - [x] Move to JSON objects instead of tuples - [x] Specify endpoints for polling for updates on single values (PollItem) **Implementation:** - [x] Table for K2V items, causal contexts - [x] Indexing mechanism and table for K2V index - [x] Make API handlers a bit more generic - [x] K2V API endpoint - [x] K2V API router - [x] ReadItem - [x] InsertItem - [x] DeleteItem - [x] PollItem - [x] ReadIndex - [x] InsertBatch - [x] ReadBatch - [x] DeleteBatch **Testing:** - [x] Just a simple Python script that does some requests to check visually that things are going right (does not contain parsing of results or assertions on returned values) - [x] Actual tests: - [x] Adapt testing framework - [x] Simple test with InsertItem + ReadItem - [x] Test with several Insert/Read/DeleteItem + ReadIndex - [x] Test all combinations of return formats for ReadItem - [x] Test with ReadBatch, InsertBatch, DeleteBatch - [x] Test with PollItem - [x] Test error codes - [ ] Fix most broken stuff - [x] test PollItem broken randomly - [x] when invalid causality tokens are given, errors should be 4xx not 5xx **Improvements:** - [x] Descending range queries - [x] Specify - [x] Implement - [x] Add test - [x] Batch updates to index counter - [x] Put K2V behind `k2v` feature flag Co-authored-by: Alex Auvolat <alex@adnab.me> Reviewed-on: https://git.deuxfleurs.fr/Deuxfleurs/garage/pulls/293 Co-authored-by: Alex <alex@adnab.me> Co-committed-by: Alex <alex@adnab.me>
2022-05-10 11:16:57 +00:00
for v in to_repair {
2022-12-14 15:08:05 +00:00
if let Err(e) = self2.repair_on_read(&who[..], v).await {
warn!("Error doing repair on read: {}", e);
}
}
});
}
First implementation of K2V (#293) **Specification:** View spec at [this URL](https://git.deuxfleurs.fr/Deuxfleurs/garage/src/branch/k2v/doc/drafts/k2v-spec.md) - [x] Specify the structure of K2V triples - [x] Specify the DVVS format used for causality detection - [x] Specify the K2V index (just a counter of number of values per partition key) - [x] Specify single-item endpoints: ReadItem, InsertItem, DeleteItem - [x] Specify index endpoint: ReadIndex - [x] Specify multi-item endpoints: InsertBatch, ReadBatch, DeleteBatch - [x] Move to JSON objects instead of tuples - [x] Specify endpoints for polling for updates on single values (PollItem) **Implementation:** - [x] Table for K2V items, causal contexts - [x] Indexing mechanism and table for K2V index - [x] Make API handlers a bit more generic - [x] K2V API endpoint - [x] K2V API router - [x] ReadItem - [x] InsertItem - [x] DeleteItem - [x] PollItem - [x] ReadIndex - [x] InsertBatch - [x] ReadBatch - [x] DeleteBatch **Testing:** - [x] Just a simple Python script that does some requests to check visually that things are going right (does not contain parsing of results or assertions on returned values) - [x] Actual tests: - [x] Adapt testing framework - [x] Simple test with InsertItem + ReadItem - [x] Test with several Insert/Read/DeleteItem + ReadIndex - [x] Test all combinations of return formats for ReadItem - [x] Test with ReadBatch, InsertBatch, DeleteBatch - [x] Test with PollItem - [x] Test error codes - [ ] Fix most broken stuff - [x] test PollItem broken randomly - [x] when invalid causality tokens are given, errors should be 4xx not 5xx **Improvements:** - [x] Descending range queries - [x] Specify - [x] Implement - [x] Add test - [x] Batch updates to index counter - [x] Put K2V behind `k2v` feature flag Co-authored-by: Alex Auvolat <alex@adnab.me> Reviewed-on: https://git.deuxfleurs.fr/Deuxfleurs/garage/pulls/293 Co-authored-by: Alex <alex@adnab.me> Co-committed-by: Alex <alex@adnab.me>
2022-05-10 11:16:57 +00:00
// At this point, the `ret` btreemap might contain more than `limit`
// items, because nodes might have returned us each `limit` items
// but for different keys. We have to take only the first `limit` items
// in this map, in the specified enumeration order, for two reasons:
// 1. To return to the user no more than the number of items that they requested
// 2. To return only items for which we have a read quorum: we do not know
// that we have a read quorum for the items after the first `limit`
// of them
let ret_vec = match enumeration_order {
EnumerationOrder::Forward => ret
.into_iter()
.take(limit)
.map(|(_k, v)| v)
.collect::<Vec<_>>(),
EnumerationOrder::Reverse => ret
.into_iter()
.rev()
.take(limit)
.map(|(_k, v)| v)
.collect::<Vec<_>>(),
};
Ok(ret_vec)
}
// =============== UTILITY FUNCTION FOR CLIENT OPERATIONS ===============
async fn repair_on_read(&self, who: &[Uuid], what: F::E) -> Result<(), Error> {
2023-01-03 13:44:47 +00:00
let what_enc = Arc::new(ByteBuf::from(what.encode()?));
2021-10-14 09:50:12 +00:00
self.system
.rpc_helper()
2020-04-18 17:21:34 +00:00
.try_call_many(
2021-10-14 09:50:12 +00:00
&self.endpoint,
2021-04-23 19:42:52 +00:00
who,
2021-05-02 21:13:08 +00:00
TableRpc::<F>::Update(vec![what_enc]),
RequestStrategy::with_priority(PRIO_NORMAL).with_quorum(who.len()),
2020-04-18 17:21:34 +00:00
)
2020-04-16 12:50:49 +00:00
.await?;
Ok(())
}
}
#[async_trait]
2023-01-03 14:08:37 +00:00
impl<F: TableSchema, R: TableReplication> EndpointHandler<TableRpc<F>> for Table<F, R> {
async fn handle(
self: &Arc<Self>,
msg: &TableRpc<F>,
_from: NodeID,
) -> Result<TableRpc<F>, Error> {
2020-04-08 21:01:49 +00:00
match msg {
2021-05-02 21:13:08 +00:00
TableRpc::ReadEntry(key, sort_key) => {
2021-03-11 15:54:15 +00:00
let value = self.data.read_entry(key, sort_key)?;
2021-05-02 21:13:08 +00:00
Ok(TableRpc::ReadEntryResponse(value))
2020-04-08 21:01:49 +00:00
}
First implementation of K2V (#293) **Specification:** View spec at [this URL](https://git.deuxfleurs.fr/Deuxfleurs/garage/src/branch/k2v/doc/drafts/k2v-spec.md) - [x] Specify the structure of K2V triples - [x] Specify the DVVS format used for causality detection - [x] Specify the K2V index (just a counter of number of values per partition key) - [x] Specify single-item endpoints: ReadItem, InsertItem, DeleteItem - [x] Specify index endpoint: ReadIndex - [x] Specify multi-item endpoints: InsertBatch, ReadBatch, DeleteBatch - [x] Move to JSON objects instead of tuples - [x] Specify endpoints for polling for updates on single values (PollItem) **Implementation:** - [x] Table for K2V items, causal contexts - [x] Indexing mechanism and table for K2V index - [x] Make API handlers a bit more generic - [x] K2V API endpoint - [x] K2V API router - [x] ReadItem - [x] InsertItem - [x] DeleteItem - [x] PollItem - [x] ReadIndex - [x] InsertBatch - [x] ReadBatch - [x] DeleteBatch **Testing:** - [x] Just a simple Python script that does some requests to check visually that things are going right (does not contain parsing of results or assertions on returned values) - [x] Actual tests: - [x] Adapt testing framework - [x] Simple test with InsertItem + ReadItem - [x] Test with several Insert/Read/DeleteItem + ReadIndex - [x] Test all combinations of return formats for ReadItem - [x] Test with ReadBatch, InsertBatch, DeleteBatch - [x] Test with PollItem - [x] Test error codes - [ ] Fix most broken stuff - [x] test PollItem broken randomly - [x] when invalid causality tokens are given, errors should be 4xx not 5xx **Improvements:** - [x] Descending range queries - [x] Specify - [x] Implement - [x] Add test - [x] Batch updates to index counter - [x] Put K2V behind `k2v` feature flag Co-authored-by: Alex Auvolat <alex@adnab.me> Reviewed-on: https://git.deuxfleurs.fr/Deuxfleurs/garage/pulls/293 Co-authored-by: Alex <alex@adnab.me> Co-committed-by: Alex <alex@adnab.me>
2022-05-10 11:16:57 +00:00
TableRpc::ReadRange {
partition,
begin_sort_key,
filter,
limit,
enumeration_order,
} => {
let values = self.data.read_range(
partition,
begin_sort_key,
filter,
*limit,
*enumeration_order,
)?;
2021-05-02 21:13:08 +00:00
Ok(TableRpc::Update(values))
}
2021-05-02 21:13:08 +00:00
TableRpc::Update(pairs) => {
2021-03-11 15:54:15 +00:00
self.data.update_many(pairs)?;
2021-05-02 21:13:08 +00:00
Ok(TableRpc::Ok)
2020-04-08 21:01:49 +00:00
}
m => Err(Error::unexpected_rpc_message(m)),
2020-04-08 21:01:49 +00:00
}
}
2020-04-08 20:00:41 +00:00
}