store: replace fjal reset by dir nuking

This commit is contained in:
nym21
2026-04-03 17:49:46 +02:00
parent 768e6870cb
commit 5bf06530ce
10 changed files with 68 additions and 424 deletions

26
Cargo.lock generated
View File

@@ -711,9 +711,7 @@ dependencies = [
"brk_types",
"byteview",
"fjall",
"rayon",
"rustc-hash",
"tempfile",
]
[[package]]
@@ -1307,9 +1305,9 @@ checksum = "5baebc0774151f905a1a2cc41989300b1e6fbb29aff0ceffa1064fdd3088d582"
[[package]]
name = "fjall"
version = "3.1.2"
version = "3.0.4"
source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "1a9530ff159bc3ad3a15da746da0f6e95375c2ac64708cbb85ec1ebd26761a84"
checksum = "0ebf22b812878dcd767879cb19e03124fd62563dce6410f96538175fba0c132d"
dependencies = [
"byteorder-lite",
"byteview",
@@ -1317,7 +1315,7 @@ dependencies = [
"flume",
"log",
"lsm-tree",
"lz4_flex",
"lz4_flex 0.11.6",
"tempfile",
"xxhash-rust",
]
@@ -2086,9 +2084,9 @@ checksum = "5e5032e24019045c762d3c0f28f5b6b8bbf38563a65908389bf7978758920897"
[[package]]
name = "lsm-tree"
version = "3.1.2"
version = "3.0.4"
source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "9d67f95fd716870329c30aaeedf87f23d426564e6ce46efa045a91444faf2a19"
checksum = "e9bfd2a6ea0c1d430c13643002f35800a87f200fc8ac4827f18a2db9d9fd0644"
dependencies = [
"byteorder-lite",
"byteview",
@@ -2096,7 +2094,7 @@ dependencies = [
"enum_dispatch",
"interval-heap",
"log",
"lz4_flex",
"lz4_flex 0.11.6",
"quick_cache",
"rustc-hash",
"self_cell",
@@ -2108,13 +2106,19 @@ dependencies = [
[[package]]
name = "lz4_flex"
version = "0.13.0"
version = "0.11.6"
source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "db9a0d582c2874f68138a16ce1867e0ffde6c0bb0a0df85e1f36d04146db488a"
checksum = "373f5eceeeab7925e0c1098212f2fbc4d416adec9d35051a6ab251e824c1854a"
dependencies = [
"twox-hash",
]
[[package]]
name = "lz4_flex"
version = "0.13.0"
source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "db9a0d582c2874f68138a16ce1867e0ffde6c0bb0a0df85e1f36d04146db488a"
[[package]]
name = "matchit"
version = "0.8.4"
@@ -3439,7 +3443,7 @@ dependencies = [
"itoa",
"libc",
"log",
"lz4_flex",
"lz4_flex 0.13.0",
"parking_lot",
"pco",
"rawdb",

View File

@@ -68,7 +68,7 @@ color-eyre = "0.6.5"
corepc-client = { package = "brk-corepc-client", version = "0.11.0", features = ["client-sync"] }
corepc-jsonrpc = { package = "brk-corepc-jsonrpc", version = "0.19.0", features = ["simple_http"], default-features = false }
derive_more = { version = "2.1.1", features = ["deref", "deref_mut"] }
fjall = "3.1.2"
fjall = "=3.0.4"
indexmap = { version = "2.13.0", features = ["serde"] }
jiff = { version = "0.2.23", features = ["perf-inline", "tz-system"], default-features = false }
owo-colors = "4.3.0"

View File

@@ -114,6 +114,18 @@ impl Indexer {
}
}
/// Fully resets the indexer by deleting stores from disk and reimporting.
/// Unlike stores.reset() which uses keyspace.clear() (leaving a journal
/// record that gets replayed on every recovery), this cleanly recreates.
fn full_reset(&mut self) -> Result<()> {
info!("Full reset...");
self.vecs.reset()?;
let stores_path = self.path.join("stores");
fs::remove_dir_all(&stores_path).ok();
self.stores = Stores::forced_import(&self.path, VERSION)?;
Ok(())
}
pub fn index(&mut self, reader: &Reader, client: &Client, exit: &Exit) -> Result<Indexes> {
self.index_(reader, client, exit, false)
}
@@ -135,9 +147,7 @@ impl Indexer {
return Ok(());
}
info!("XOR bytes changed, full reset...");
self.vecs.reset()?;
self.stores.reset()?;
self.full_reset()?;
fs::write(self.path.join("xor.dat"), *current)?;
@@ -179,8 +189,7 @@ impl Indexer {
}
None => {
info!("Data inconsistency detected, resetting indexer...");
self.vecs.reset()?;
self.stores.reset()?;
self.full_reset()?;
(Indexes::default(), None)
}
}
@@ -308,13 +317,35 @@ impl Indexer {
drop(readers);
let lock = exit.lock();
self.stores.commit(indexes.height)?;
let tasks = self.stores.take_all_pending_ingests(indexes.height)?;
self.vecs.stamped_write(indexes.height)?;
let fjall_db = self.stores.db.clone();
self.vecs.db.run_bg(move |db| {
let _lock = lock;
sleep(Duration::from_secs(5));
info!("Exporting...");
let i = Instant::now();
if !tasks.is_empty() {
let i = Instant::now();
for task in tasks {
task().map_err(vecdb::RawDBError::other)?;
}
debug!("Stores committed in {:?}", i.elapsed());
let i = Instant::now();
fjall_db
.persist(PersistMode::SyncData)
.map_err(RawDBError::other)?;
debug!("Stores persisted in {:?}", i.elapsed());
}
db.compact()?;
info!("Exported in {:?}", i.elapsed());
Ok(())
});

View File

@@ -116,13 +116,6 @@ impl Stores {
)?,
};
debug!(
"Stores imported: txid_prefix empty={}, blockhash empty={}, keyspace_count={}",
stores.txid_prefix_to_tx_index.is_empty()?,
stores.blockhash_prefix_to_height.is_empty()?,
database.keyspace_count(),
);
Ok(stores)
}
@@ -409,17 +402,5 @@ impl Stores {
}
}
pub fn reset(&mut self) -> Result<()> {
info!("Resetting stores...");
// Clear all stores (both in-memory buffers and on-disk keyspaces)
self.par_iter_any_mut()
.try_for_each(|store| store.reset())?;
// Persist the cleared state
self.db.persist(PersistMode::SyncAll)?;
Ok(())
}
}

View File

@@ -26,8 +26,9 @@ pub fn init(path: Option<&Path>) -> io::Result<()> {
let level = std::env::var("LOG").unwrap_or_else(|_| DEFAULT_LEVEL.to_string());
let directives = std::env::var("RUST_LOG").unwrap_or_else(|_| {
// fjall=off,lsm_tree=off
format!(
"{level},bitcoin=off,bitcoincore_rpc=off,corepc=off,fjall=off,brk_fjall=off,lsm_tree=off,brk_rolldown=off,rolldown=off,tracing=off,aide=off,rustls=off,notify=off,oxc_resolver=off,tower_http=off"
"{level},bitcoin=off,bitcoincore_rpc=off,corepc=off,tracing=off,aide=off,tower_http=off"
)
});

View File

@@ -16,7 +16,3 @@ brk_types = { workspace = true }
byteview = { workspace = true }
fjall = { workspace = true }
rustc-hash = { workspace = true }
[dev-dependencies]
rayon.workspace = true
tempfile = "3.27.0"

View File

@@ -12,5 +12,4 @@ pub trait AnyStore: Send + Sync {
fn export_meta_if_needed(&mut self, height: Height) -> Result<()>;
fn keyspace(&self) -> &Keyspace;
fn commit(&mut self, height: Height) -> Result<()>;
fn reset(&mut self) -> Result<()>;
}

View File

@@ -365,14 +365,4 @@ where
Ok(())
}
fn reset(&mut self) -> Result<()> {
self.meta.reset()?;
self.puts = FxHashMap::default();
self.dels = FxHashSet::default();
for cache in &mut self.caches {
*cache = FxHashMap::default();
}
self.keyspace.clear()?;
Ok(())
}
}

View File

@@ -1,365 +0,0 @@
//! Minimal reproduction: data written via start_ingestion is lost after close+reopen.
//!
//! This mimics what brk does:
//! 1. Open database with manual_journal_persist
//! 2. Create a keyspace (Kind::Recent config)
//! 3. Use start_ingestion to bulk-write data
//! 4. Call persist(SyncData)
//! 5. Drop the database
//! 6. Reopen
//! 7. Check if data survived
use brk_store::{Kind, Mode, Store};
use brk_types::{Height, TxIndex, TxidPrefix, Version};
use fjall::{Database, KeyspaceCreateOptions, PersistMode};
fn open_db(path: &std::path::Path) -> Database {
Database::builder(path.join("fjall"))
.cache_size(64 * 1024 * 1024)
.open()
.unwrap()
}
fn open_keyspace(db: &Database) -> fjall::Keyspace {
db.keyspace("test_keyspace", || {
KeyspaceCreateOptions::default()
.manual_journal_persist(true)
.expect_point_read_hits(true)
})
.unwrap()
}
#[test]
fn ingestion_survives_close_reopen() {
let dir = tempfile::tempdir().unwrap();
let path = dir.path();
// Phase 1: write data via ingestion, then close
{
let db = open_db(path);
let ks = open_keyspace(&db);
let mut ingestion = ks.start_ingestion().unwrap();
for i in 0u64..1000 {
ingestion
.write(i.to_be_bytes(), i.to_be_bytes())
.unwrap();
}
ingestion.finish().unwrap();
// Verify data is readable before close
assert!(!ks.is_empty().unwrap(), "keyspace should have data before close");
assert!(ks.get(0u64.to_be_bytes()).unwrap().is_some(), "key 0 should exist before close");
db.persist(PersistMode::SyncData).unwrap();
// db + ks dropped here
}
// Phase 2: reopen and check
{
let db = open_db(path);
let ks = open_keyspace(&db);
assert!(
!ks.is_empty().unwrap(),
"BUG: keyspace is empty after close+reopen — ingested data lost"
);
assert!(
ks.get(0u64.to_be_bytes()).unwrap().is_some(),
"BUG: key 0 missing after close+reopen"
);
assert!(
ks.get(999u64.to_be_bytes()).unwrap().is_some(),
"BUG: key 999 missing after close+reopen"
);
}
}
/// Same test but with a keyspace clone (mimics take_pending_ingest capturing keyspace.clone())
#[test]
fn ingestion_via_cloned_keyspace_survives_close_reopen() {
let dir = tempfile::tempdir().unwrap();
let path = dir.path();
{
let db = open_db(path);
let ks = open_keyspace(&db);
// Clone the keyspace (like take_pending_ingest does)
let ks_clone = ks.clone();
let mut ingestion = ks_clone.start_ingestion().unwrap();
for i in 0u64..1000 {
ingestion
.write(i.to_be_bytes(), i.to_be_bytes())
.unwrap();
}
ingestion.finish().unwrap();
// Clone used for persist (like fjall_db.persist in bg task)
let db_clone = db.clone();
db_clone.persist(PersistMode::SyncData).unwrap();
// Drop order mimics Indexer: ks_clone dropped first, then db_clone, then ks, then db
drop(ks_clone);
drop(db_clone);
drop(ks);
drop(db);
}
{
let db = open_db(path);
let ks = open_keyspace(&db);
assert!(
!ks.is_empty().unwrap(),
"BUG: keyspace is empty after close+reopen — cloned ingestion data lost"
);
assert!(
ks.get(500u64.to_be_bytes()).unwrap().is_some(),
"BUG: key 500 missing after close+reopen (cloned keyspace path)"
);
}
}
/// Mimics brk at scale: 20+ keyspaces, parallel intermediate commits (like par_iter_any_mut),
/// hundreds of batches, large data, bg thread ingest, drop-db-before-keyspaces order.
#[test]
fn many_keyspaces_parallel_commits_bg_ingest() {
use rayon::prelude::*;
let dir = tempfile::tempdir().unwrap();
let path = dir.path();
const NUM_KEYSPACES: usize = 25;
const INTERMEDIATE_BATCHES: u64 = 500;
const KEYS_PER_BATCH: u64 = 10_000;
const BG_KEYS_PER_KS: u64 = 10_000;
{
let db = open_db(path);
let keyspaces: Vec<fjall::Keyspace> = (0..NUM_KEYSPACES)
.map(|i| {
db.keyspace(&format!("ks_{i}"), || {
let mut opts = KeyspaceCreateOptions::default()
.manual_journal_persist(true);
// Mix configs like brk does (Kind::Recent vs Kind::Random vs Kind::Vec)
if i % 3 == 0 {
opts = opts.expect_point_read_hits(true);
}
opts
})
.unwrap()
})
.collect();
// Intermediate commits — PARALLEL across keyspaces (like par_iter_any_mut)
for batch in 0..INTERMEDIATE_BATCHES {
keyspaces.par_iter().for_each(|ks| {
let start = batch * KEYS_PER_BATCH;
let end = start + KEYS_PER_BATCH;
let mut ing = ks.start_ingestion().unwrap();
for i in start..end {
ing.write(i.to_be_bytes(), i.to_be_bytes()).unwrap();
}
ing.finish().unwrap();
});
db.persist(PersistMode::SyncData).unwrap();
}
let total_intermediate = INTERMEDIATE_BATCHES * KEYS_PER_BATCH;
eprintln!("Wrote {total_intermediate} keys/ks × {NUM_KEYSPACES} keyspaces in {INTERMEDIATE_BATCHES} parallel batches");
// take_pending_ingest: clone each keyspace + db, run on bg thread SEQUENTIALLY
let ks_clones: Vec<_> = keyspaces.iter().map(|ks| ks.clone()).collect();
let db_clone = db.clone();
let handle = std::thread::spawn(move || {
std::thread::sleep(std::time::Duration::from_millis(500));
// Sequential ingestion per keyspace (like `for task in tasks { task()?; }`)
for ks_clone in &ks_clones {
let start = total_intermediate;
let end = start + BG_KEYS_PER_KS;
let mut ing = ks_clone.start_ingestion().unwrap();
for i in start..end {
ing.write(i.to_be_bytes(), i.to_be_bytes()).unwrap();
}
ing.finish().unwrap();
}
db_clone.persist(PersistMode::SyncData).unwrap();
});
// sync_bg_tasks
handle.join().unwrap();
// Stores drop order: db first, then keyspaces (struct field order)
drop(db);
drop(keyspaces);
}
// Reopen and verify
{
let db = open_db(path);
let total_intermediate = INTERMEDIATE_BATCHES * KEYS_PER_BATCH;
for i in 0..NUM_KEYSPACES {
let ks = db
.keyspace(&format!("ks_{i}"), || {
KeyspaceCreateOptions::default().manual_journal_persist(true)
})
.unwrap();
assert!(
!ks.is_empty().unwrap(),
"BUG: ks_{i} is empty after reopen"
);
// Intermediate data
assert!(
ks.get(0u64.to_be_bytes()).unwrap().is_some(),
"BUG: ks_{i} key 0 missing"
);
assert!(
ks.get((total_intermediate - 1).to_be_bytes()).unwrap().is_some(),
"BUG: ks_{i} key {} missing", total_intermediate - 1
);
// Bg task data
let bg_mid = total_intermediate + BG_KEYS_PER_KS / 2;
assert!(
ks.get(bg_mid.to_be_bytes()).unwrap().is_some(),
"BUG: ks_{i} key {bg_mid} (bg) missing"
);
// Spot checks across the full range
for check in [1u64, 100, 1_000, 10_000, 100_000, 1_000_000, 4_999_999] {
if check < total_intermediate + BG_KEYS_PER_KS {
assert!(
ks.get(check.to_be_bytes()).unwrap().is_some(),
"BUG: ks_{i} key {check} missing"
);
}
}
}
eprintln!("All {NUM_KEYSPACES} keyspaces verified after reopen");
}
}
/// Uses the ACTUAL brk Store<TxidPrefix, TxIndex> type with commit + take_pending_ingest.
/// This exercises the exact code path that brk uses.
#[test]
fn actual_store_commit_then_take_pending_ingest() -> Result<(), Box<dyn std::error::Error>> {
let dir = tempfile::tempdir().unwrap();
let path = dir.path();
let stores_path = path.join("stores");
std::fs::create_dir_all(&stores_path).unwrap();
let version = Version::new(29); // MAJOR_FJALL_VERSION(3) + VERSION(26)
{
let db = brk_store::open_database(&stores_path).unwrap();
let mut store: Store<TxidPrefix, TxIndex> = Store::import_cached(
&db,
&stores_path,
"txid_prefix_to_tx_index",
version,
Mode::PushOnly,
Kind::Recent,
5,
)
.unwrap();
// Simulate intermediate commits (like Stores::commit every 1000 blocks)
for batch in 0u64..500 {
for i in (batch * 1000)..((batch + 1) * 1000) {
let prefix = TxidPrefix::from(byteview::ByteView::from(i.to_be_bytes()));
let tx_index = TxIndex::from(i as usize);
store.insert(prefix, tx_index);
}
// AnyStore::commit
brk_store::AnyStore::commit(&mut store, Height::from(batch as u32))?;
db.persist(PersistMode::SyncData).unwrap();
}
let total_intermediate = 500_000u64;
// Verify before take_pending_ingest
let prefix_0 = TxidPrefix::from(byteview::ByteView::from(0u64.to_be_bytes()));
assert!(store.get(&prefix_0).unwrap().is_some(), "key 0 should exist before take");
// Simulate take_pending_ingest: add more data, then take
for i in total_intermediate..(total_intermediate + 5_000) {
let prefix = TxidPrefix::from(byteview::ByteView::from(i.to_be_bytes()));
let tx_index = TxIndex::from(i as usize);
store.insert(prefix, tx_index);
}
let task = store
.take_pending_ingest(Height::from(943425u32))
.unwrap();
// Simulate bg thread
let db_clone = db.clone();
let handle = std::thread::spawn(move || {
if let Some(task) = task {
task().unwrap();
}
db_clone.persist(PersistMode::SyncData).unwrap();
});
handle.join().unwrap();
// Drop order: db first, then store (like Stores struct)
drop(db);
drop(store);
}
// Reopen and verify
{
let db = brk_store::open_database(&stores_path).unwrap();
let store: Store<TxidPrefix, TxIndex> = Store::import_cached(
&db,
&stores_path,
"txid_prefix_to_tx_index",
version,
Mode::PushOnly,
Kind::Recent,
5,
)
.unwrap();
assert!(
!store.is_empty().unwrap(),
"BUG: store is empty after reopen"
);
// Check intermediate data
let prefix_0 = TxidPrefix::from(byteview::ByteView::from(0u64.to_be_bytes()));
assert!(
store.get(&prefix_0).unwrap().is_some(),
"BUG: key 0 (intermediate) missing after reopen"
);
let prefix_mid = TxidPrefix::from(byteview::ByteView::from(250_000u64.to_be_bytes()));
assert!(
store.get(&prefix_mid).unwrap().is_some(),
"BUG: key 250000 (intermediate) missing after reopen"
);
// Check bg task data
let prefix_bg = TxidPrefix::from(byteview::ByteView::from(502_000u64.to_be_bytes()));
assert!(
store.get(&prefix_bg).unwrap().is_some(),
"BUG: key 502000 (bg task) missing after reopen"
);
}
Ok(())
}

View File

@@ -182,11 +182,11 @@ function renderDetails(block) {
["Miner Names", extras.pool.minerNames || "N/A"],
["Reward", `${(extras.reward / 1e8).toFixed(8)} BTC`],
["Total Fees", `${(extras.totalFees / 1e8).toFixed(8)} BTC`],
["Median Fee Rate", `${extras.medianFee.toFixed(2)} sat/vB`],
["Avg Fee Rate", `${extras.avgFeeRate.toFixed(2)} sat/vB`],
["Median Fee Rate", `${formatFeeRate(extras.medianFee)} sat/vB`],
["Avg Fee Rate", `${formatFeeRate(extras.avgFeeRate)} sat/vB`],
["Avg Fee", `${extras.avgFee.toLocaleString()} sat`],
["Median Fee", `${extras.medianFeeAmt.toLocaleString()} sat`],
["Fee Range", extras.feeRange.map((f) => f.toFixed(1)).join(", ") + " sat/vB"],
["Fee Range", extras.feeRange.map((f) => formatFeeRate(f)).join(", ") + " sat/vB"],
["Fee Percentiles", extras.feePercentiles.map((f) => f.toLocaleString()).join(", ") + " sat"],
["Avg Tx Size", `${extras.avgTxSize.toLocaleString()} B`],
["Virtual Size", `${extras.virtualSize.toLocaleString()} vB`],
@@ -222,6 +222,13 @@ function renderDetails(block) {
}
}
/** @param {number} rate */
function formatFeeRate(rate) {
if (rate >= 100) return Math.round(rate).toLocaleString();
if (rate >= 10) return rate.toFixed(1);
return rate.toFixed(2);
}
/** @param {number} height */
function createHeightElement(height) {
const container = window.document.createElement("span");
@@ -257,18 +264,18 @@ function createBlockCube(block) {
const feeRange = extras ? extras.feeRange : [0, 0, 0, 0, 0, 0, 0];
const averageFeeElement = window.document.createElement("p");
feesElement.append(averageFeeElement);
averageFeeElement.innerHTML = `~${Number(medianFee).toFixed(2)}`;
averageFeeElement.innerHTML = `~${formatFeeRate(medianFee)}`;
const feeRangeElement = window.document.createElement("p");
feesElement.append(feeRangeElement);
const minFeeElement = window.document.createElement("span");
minFeeElement.innerHTML = `${Number(feeRange[0]).toFixed(2)}`;
minFeeElement.innerHTML = formatFeeRate(feeRange[0]);
feeRangeElement.append(minFeeElement);
const dashElement = window.document.createElement("span");
dashElement.style.opacity = "0.5";
dashElement.innerHTML = `-`;
feeRangeElement.append(dashElement);
const maxFeeElement = window.document.createElement("span");
maxFeeElement.innerHTML = `${Number(feeRange[6]).toFixed(1)}`;
maxFeeElement.innerHTML = formatFeeRate(feeRange[6]);
feeRangeElement.append(maxFeeElement);
const feeUnitElement = window.document.createElement("p");
feesElement.append(feeUnitElement);