Remove schema migrations for v28 and earlier (#9031)

With LH v8.1.3 supporting Fulu-on-Gnosis, we no longer need these DB migrations. All Lighthouse nodes running in prod will soon be updated to LH v8.0.0+ and schema v28+.

This PR helps with Gloas fork choice changes, by allowing us to avoid updating old schema migrations when adding V29 for Gloas:

- https://github.com/sigp/lighthouse/pull/9025


  


Co-Authored-By: Michael Sproul <michael@sigmaprime.io>
This commit is contained in:
Michael Sproul
2026-03-26 13:10:34 +11:00
committed by GitHub
parent c7055b604f
commit bd34bb1430
19 changed files with 23 additions and 1578 deletions

View File

@@ -231,35 +231,6 @@ where
}
}
/// Restore `Self` from a previously-generated `PersistedForkChoiceStore`.
///
/// DEPRECATED. Can be deleted once migrations no longer require it.
pub fn from_persisted_v17(
persisted: PersistedForkChoiceStoreV17,
justified_state_root: Hash256,
unrealized_justified_state_root: Hash256,
store: Arc<HotColdDB<E, Hot, Cold>>,
) -> Result<Self, Error> {
let justified_balances =
JustifiedBalances::from_effective_balances(persisted.justified_balances)?;
Ok(Self {
store,
balances_cache: <_>::default(),
time: persisted.time,
finalized_checkpoint: persisted.finalized_checkpoint,
justified_checkpoint: persisted.justified_checkpoint,
justified_balances,
justified_state_root,
unrealized_justified_checkpoint: persisted.unrealized_justified_checkpoint,
unrealized_justified_state_root,
unrealized_finalized_checkpoint: persisted.unrealized_finalized_checkpoint,
proposer_boost_root: persisted.proposer_boost_root,
equivocating_indices: persisted.equivocating_indices,
_phantom: PhantomData,
})
}
/// Restore `Self` from a previously-generated `PersistedForkChoiceStore`.
pub fn from_persisted(
persisted: PersistedForkChoiceStore,
@@ -411,45 +382,15 @@ where
pub type PersistedForkChoiceStore = PersistedForkChoiceStoreV28;
/// A container which allows persisting the `BeaconForkChoiceStore` to the on-disk database.
#[superstruct(
variants(V17, V28),
variant_attributes(derive(Encode, Decode)),
no_enum
)]
#[superstruct(variants(V28), variant_attributes(derive(Encode, Decode)), no_enum)]
pub struct PersistedForkChoiceStore {
/// The balances cache was removed from disk storage in schema V28.
#[superstruct(only(V17))]
pub balances_cache: BalancesCacheV8,
pub time: Slot,
pub finalized_checkpoint: Checkpoint,
pub justified_checkpoint: Checkpoint,
/// The justified balances were removed from disk storage in schema V28.
#[superstruct(only(V17))]
pub justified_balances: Vec<u64>,
/// The justified state root is stored so that it can be used to load the justified balances.
#[superstruct(only(V28))]
pub justified_state_root: Hash256,
pub unrealized_justified_checkpoint: Checkpoint,
#[superstruct(only(V28))]
pub unrealized_justified_state_root: Hash256,
pub unrealized_finalized_checkpoint: Checkpoint,
pub proposer_boost_root: Hash256,
pub equivocating_indices: BTreeSet<u64>,
}
// Convert V28 to V17 by adding balances and removing justified state roots.
impl From<(PersistedForkChoiceStoreV28, JustifiedBalances)> for PersistedForkChoiceStoreV17 {
fn from((v28, balances): (PersistedForkChoiceStoreV28, JustifiedBalances)) -> Self {
Self {
balances_cache: Default::default(),
time: v28.time,
finalized_checkpoint: v28.finalized_checkpoint,
justified_checkpoint: v28.justified_checkpoint,
justified_balances: balances.effective_balances,
unrealized_justified_checkpoint: v28.unrealized_justified_checkpoint,
unrealized_finalized_checkpoint: v28.unrealized_finalized_checkpoint,
proposer_boost_root: v28.proposer_boost_root,
equivocating_indices: v28.equivocating_indices,
}
}
}

View File

@@ -75,7 +75,7 @@ pub use self::errors::{BeaconChainError, BlockProductionError};
pub use self::historical_blocks::HistoricalBlockError;
pub use attestation_verification::Error as AttestationError;
pub use beacon_fork_choice_store::{
BeaconForkChoiceStore, Error as ForkChoiceStoreError, PersistedForkChoiceStoreV17,
BeaconForkChoiceStore, Error as ForkChoiceStoreError, PersistedForkChoiceStore,
PersistedForkChoiceStoreV28,
};
pub use block_verification::{

View File

@@ -1,52 +1,19 @@
use crate::{
beacon_fork_choice_store::{PersistedForkChoiceStoreV17, PersistedForkChoiceStoreV28},
metrics,
};
use crate::{beacon_fork_choice_store::PersistedForkChoiceStoreV28, metrics};
use ssz::{Decode, Encode};
use ssz_derive::{Decode, Encode};
use store::{DBColumn, Error, KeyValueStoreOp, StoreConfig, StoreItem};
use store::{DBColumn, Error, KeyValueStoreOp, StoreConfig};
use superstruct::superstruct;
use types::Hash256;
// If adding a new version you should update this type alias and fix the breakages.
pub type PersistedForkChoice = PersistedForkChoiceV28;
#[superstruct(
variants(V17, V28),
variant_attributes(derive(Encode, Decode)),
no_enum
)]
#[superstruct(variants(V28), variant_attributes(derive(Encode, Decode)), no_enum)]
pub struct PersistedForkChoice {
#[superstruct(only(V17))]
pub fork_choice_v17: fork_choice::PersistedForkChoiceV17,
#[superstruct(only(V28))]
pub fork_choice: fork_choice::PersistedForkChoiceV28,
#[superstruct(only(V17))]
pub fork_choice_store_v17: PersistedForkChoiceStoreV17,
#[superstruct(only(V28))]
pub fork_choice_store: PersistedForkChoiceStoreV28,
}
macro_rules! impl_store_item {
($type:ty) => {
impl StoreItem for $type {
fn db_column() -> DBColumn {
DBColumn::ForkChoice
}
fn as_store_bytes(&self) -> Vec<u8> {
self.as_ssz_bytes()
}
fn from_store_bytes(bytes: &[u8]) -> std::result::Result<Self, Error> {
Self::from_ssz_bytes(bytes).map_err(Into::into)
}
}
};
}
impl_store_item!(PersistedForkChoiceV17);
impl PersistedForkChoiceV28 {
pub fn from_bytes(bytes: &[u8], store_config: &StoreConfig) -> Result<Self, Error> {
let decompressed_bytes = store_config

View File

@@ -1,11 +1,4 @@
//! Utilities for managing database schema changes.
mod migration_schema_v23;
mod migration_schema_v24;
mod migration_schema_v25;
mod migration_schema_v26;
mod migration_schema_v27;
mod migration_schema_v28;
use crate::beacon_chain::BeaconChainTypes;
use std::sync::Arc;
use store::Error as StoreError;
@@ -13,81 +6,17 @@ use store::hot_cold_store::{HotColdDB, HotColdDBError};
use store::metadata::{CURRENT_SCHEMA_VERSION, SchemaVersion};
/// Migrate the database from one schema version to another, applying all requisite mutations.
///
/// All migrations for schema versions up to and including v28 have been removed. Nodes on live
/// networks are already running v28, so only the current version check remains.
pub fn migrate_schema<T: BeaconChainTypes>(
db: Arc<HotColdDB<T::EthSpec, T::HotStore, T::ColdStore>>,
_db: Arc<HotColdDB<T::EthSpec, T::HotStore, T::ColdStore>>,
from: SchemaVersion,
to: SchemaVersion,
) -> Result<(), StoreError> {
match (from, to) {
// Migrating from the current schema version to itself is always OK, a no-op.
(_, _) if from == to && to == CURRENT_SCHEMA_VERSION => Ok(()),
// Upgrade across multiple versions by recursively migrating one step at a time.
(_, _) if from.as_u64() + 1 < to.as_u64() => {
let next = SchemaVersion(from.as_u64() + 1);
migrate_schema::<T>(db.clone(), from, next)?;
migrate_schema::<T>(db, next, to)
}
// Downgrade across multiple versions by recursively migrating one step at a time.
(_, _) if to.as_u64() + 1 < from.as_u64() => {
let next = SchemaVersion(from.as_u64() - 1);
migrate_schema::<T>(db.clone(), from, next)?;
migrate_schema::<T>(db, next, to)
}
//
// Migrations from before SchemaVersion(22) are deprecated.
//
(SchemaVersion(22), SchemaVersion(23)) => {
let ops = migration_schema_v23::upgrade_to_v23::<T>(db.clone())?;
db.store_schema_version_atomically(to, ops)
}
(SchemaVersion(23), SchemaVersion(22)) => {
let ops = migration_schema_v23::downgrade_from_v23::<T>(db.clone())?;
db.store_schema_version_atomically(to, ops)
}
(SchemaVersion(23), SchemaVersion(24)) => {
let ops = migration_schema_v24::upgrade_to_v24::<T>(db.clone())?;
db.store_schema_version_atomically(to, ops)
}
(SchemaVersion(24), SchemaVersion(23)) => {
let ops = migration_schema_v24::downgrade_from_v24::<T>(db.clone())?;
db.store_schema_version_atomically(to, ops)
}
(SchemaVersion(24), SchemaVersion(25)) => {
let ops = migration_schema_v25::upgrade_to_v25()?;
db.store_schema_version_atomically(to, ops)
}
(SchemaVersion(25), SchemaVersion(24)) => {
let ops = migration_schema_v25::downgrade_from_v25()?;
db.store_schema_version_atomically(to, ops)
}
(SchemaVersion(25), SchemaVersion(26)) => {
let ops = migration_schema_v26::upgrade_to_v26::<T>(db.clone())?;
db.store_schema_version_atomically(to, ops)
}
(SchemaVersion(26), SchemaVersion(25)) => {
let ops = migration_schema_v26::downgrade_from_v26::<T>(db.clone())?;
db.store_schema_version_atomically(to, ops)
}
(SchemaVersion(26), SchemaVersion(27)) => {
// This migration updates the blobs db. The schema version
// is bumped inside upgrade_to_v27.
migration_schema_v27::upgrade_to_v27::<T>(db.clone())
}
(SchemaVersion(27), SchemaVersion(26)) => {
// Downgrading is essentially a no-op and is only possible
// if peer das isn't scheduled.
migration_schema_v27::downgrade_from_v27::<T>(db.clone())?;
db.store_schema_version_atomically(to, vec![])
}
(SchemaVersion(27), SchemaVersion(28)) => {
let ops = migration_schema_v28::upgrade_to_v28::<T>(db.clone())?;
db.store_schema_version_atomically(to, ops)
}
(SchemaVersion(28), SchemaVersion(27)) => {
let ops = migration_schema_v28::downgrade_from_v28::<T>(db.clone())?;
db.store_schema_version_atomically(to, ops)
}
// Anything else is an error.
(_, _) => Err(HotColdDBError::UnsupportedSchemaVersion {
target_version: to,

View File

@@ -1,180 +0,0 @@
use crate::BeaconForkChoiceStore;
use crate::beacon_chain::BeaconChainTypes;
use crate::persisted_fork_choice::PersistedForkChoiceV17;
use crate::schema_change::StoreError;
use crate::test_utils::{BEACON_CHAIN_DB_KEY, FORK_CHOICE_DB_KEY, PersistedBeaconChain};
use fork_choice::{ForkChoice, ResetPayloadStatuses};
use ssz::{Decode, Encode};
use ssz_derive::{Decode, Encode};
use std::sync::Arc;
use store::{DBColumn, Error, HotColdDB, KeyValueStore, KeyValueStoreOp, StoreItem};
use tracing::{debug, info};
use types::{Hash256, Slot};
/// Dummy value to use for the canonical head block root, see below.
pub const DUMMY_CANONICAL_HEAD_BLOCK_ROOT: Hash256 = Hash256::repeat_byte(0xff);
pub fn upgrade_to_v23<T: BeaconChainTypes>(
db: Arc<HotColdDB<T::EthSpec, T::HotStore, T::ColdStore>>,
) -> Result<Vec<KeyValueStoreOp>, Error> {
info!("Upgrading DB schema from v22 to v23");
// 1) Set the head-tracker to empty
let Some(persisted_beacon_chain_v22) =
db.get_item::<PersistedBeaconChainV22>(&BEACON_CHAIN_DB_KEY)?
else {
return Err(Error::MigrationError(
"No persisted beacon chain found in DB. Datadir could be incorrect or DB could be corrupt".to_string()
));
};
let persisted_beacon_chain = PersistedBeaconChain {
genesis_block_root: persisted_beacon_chain_v22.genesis_block_root,
};
let mut ops = vec![persisted_beacon_chain.as_kv_store_op(BEACON_CHAIN_DB_KEY)];
// 2) Wipe out all state temporary flags. While un-used in V23, if there's a rollback we could
// end-up with an inconsistent DB.
for state_root_result in db
.hot_db
.iter_column_keys::<Hash256>(DBColumn::BeaconStateTemporary)
{
let state_root = state_root_result?;
debug!(
?state_root,
"Deleting temporary state on v23 schema migration"
);
ops.push(KeyValueStoreOp::DeleteKey(
DBColumn::BeaconStateTemporary,
state_root.as_slice().to_vec(),
));
// We also delete the temporary states themselves. Although there are known issue with
// temporary states and this could lead to DB corruption, we will only corrupt the DB in
// cases where the DB would be corrupted by restarting on v7.0.x. We consider these DBs
// "too far gone". Deleting here has the advantage of not generating warnings about
// disjoint state DAGs in the v24 upgrade, or the first pruning after migration.
ops.push(KeyValueStoreOp::DeleteKey(
DBColumn::BeaconState,
state_root.as_slice().to_vec(),
));
ops.push(KeyValueStoreOp::DeleteKey(
DBColumn::BeaconStateSummary,
state_root.as_slice().to_vec(),
));
}
Ok(ops)
}
pub fn downgrade_from_v23<T: BeaconChainTypes>(
db: Arc<HotColdDB<T::EthSpec, T::HotStore, T::ColdStore>>,
) -> Result<Vec<KeyValueStoreOp>, Error> {
let Some(persisted_beacon_chain) = db.get_item::<PersistedBeaconChain>(&BEACON_CHAIN_DB_KEY)?
else {
// The `PersistedBeaconChain` must exist if fork choice exists.
return Err(Error::MigrationError(
"No persisted beacon chain found in DB. Datadir could be incorrect or DB could be corrupt".to_string(),
));
};
// Recreate head-tracker from fork choice.
let Some(persisted_fork_choice) = db.get_item::<PersistedForkChoiceV17>(&FORK_CHOICE_DB_KEY)?
else {
// Fork choice should exist if the database exists.
return Err(Error::MigrationError(
"No fork choice found in DB".to_string(),
));
};
// We use dummy roots for the justified states because we can source the balances from the v17
// persited fork choice. The justified state root isn't required to look up the justified state's
// balances (as it would be in V28). This fork choice object with corrupt state roots SHOULD NOT
// be written to disk.
let dummy_justified_state_root = Hash256::repeat_byte(0x66);
let dummy_unrealized_justified_state_root = Hash256::repeat_byte(0x77);
let fc_store = BeaconForkChoiceStore::from_persisted_v17(
persisted_fork_choice.fork_choice_store_v17,
dummy_justified_state_root,
dummy_unrealized_justified_state_root,
db.clone(),
)
.map_err(|e| {
Error::MigrationError(format!(
"Error loading fork choice store from persisted: {e:?}"
))
})?;
// Doesn't matter what policy we use for invalid payloads, as our head calculation just
// considers descent from finalization.
let reset_payload_statuses = ResetPayloadStatuses::OnlyWithInvalidPayload;
let fork_choice = ForkChoice::from_persisted(
persisted_fork_choice.fork_choice_v17.try_into()?,
reset_payload_statuses,
fc_store,
&db.spec,
)
.map_err(|e| {
Error::MigrationError(format!("Error loading fork choice from persisted: {e:?}"))
})?;
let heads = fork_choice
.proto_array()
.heads_descended_from_finalization::<T::EthSpec>(fork_choice.finalized_checkpoint());
let head_roots = heads.iter().map(|node| node.root).collect();
let head_slots = heads.iter().map(|node| node.slot).collect();
let persisted_beacon_chain_v22 = PersistedBeaconChainV22 {
_canonical_head_block_root: DUMMY_CANONICAL_HEAD_BLOCK_ROOT,
genesis_block_root: persisted_beacon_chain.genesis_block_root,
ssz_head_tracker: SszHeadTracker {
roots: head_roots,
slots: head_slots,
},
};
let ops = vec![persisted_beacon_chain_v22.as_kv_store_op(BEACON_CHAIN_DB_KEY)];
Ok(ops)
}
/// Helper struct that is used to encode/decode the state of the `HeadTracker` as SSZ bytes.
///
/// This is used when persisting the state of the `BeaconChain` to disk.
#[derive(Encode, Decode, Clone)]
pub struct SszHeadTracker {
roots: Vec<Hash256>,
slots: Vec<Slot>,
}
#[derive(Clone, Encode, Decode)]
pub struct PersistedBeaconChainV22 {
/// This value is ignored to resolve the issue described here:
///
/// https://github.com/sigp/lighthouse/pull/1639
///
/// Its removal is tracked here:
///
/// https://github.com/sigp/lighthouse/issues/1784
pub _canonical_head_block_root: Hash256,
pub genesis_block_root: Hash256,
/// DEPRECATED
pub ssz_head_tracker: SszHeadTracker,
}
impl StoreItem for PersistedBeaconChainV22 {
fn db_column() -> DBColumn {
DBColumn::BeaconChain
}
fn as_store_bytes(&self) -> Vec<u8> {
self.as_ssz_bytes()
}
fn from_store_bytes(bytes: &[u8]) -> Result<Self, StoreError> {
Self::from_ssz_bytes(bytes).map_err(Into::into)
}
}

View File

@@ -1,607 +0,0 @@
use crate::{
beacon_chain::BeaconChainTypes,
summaries_dag::{DAGStateSummary, DAGStateSummaryV22, StateSummariesDAG},
};
use ssz::{Decode, DecodeError, Encode};
use ssz_derive::Encode;
use std::{
sync::Arc,
time::{Duration, Instant},
};
use store::{
DBColumn, Error, HotColdDB, HotStateSummary, KeyValueStore, KeyValueStoreOp, StoreItem,
hdiff::StorageStrategy,
hot_cold_store::{HotStateSummaryV22, OptionalDiffBaseState},
};
use tracing::{debug, info, warn};
use types::{
BeaconState, CACHED_EPOCHS, ChainSpec, Checkpoint, CommitteeCache, EthSpec, Hash256, Slot,
execution::StatePayloadStatus,
};
/// We stopped using the pruning checkpoint in schema v23 but never explicitly deleted it.
///
/// We delete it as part of the v24 migration.
pub const PRUNING_CHECKPOINT_KEY: Hash256 = Hash256::repeat_byte(3);
pub fn store_full_state_v22<E: EthSpec>(
state_root: &Hash256,
state: &BeaconState<E>,
ops: &mut Vec<KeyValueStoreOp>,
) -> Result<(), Error> {
let bytes = StorageContainer::new(state).as_ssz_bytes();
ops.push(KeyValueStoreOp::PutKeyValue(
DBColumn::BeaconState,
state_root.as_slice().to_vec(),
bytes,
));
Ok(())
}
/// Fetch a V22 state from the database either as a full state or using block replay.
pub fn get_state_v22<T: BeaconChainTypes>(
db: &Arc<HotColdDB<T::EthSpec, T::HotStore, T::ColdStore>>,
state_root: &Hash256,
spec: &ChainSpec,
) -> Result<Option<BeaconState<T::EthSpec>>, Error> {
let Some(summary) = db.get_item::<HotStateSummaryV22>(state_root)? else {
return Ok(None);
};
let Some(base_state) =
get_full_state_v22(&db.hot_db, &summary.epoch_boundary_state_root, spec)?
else {
return Ok(None);
};
// Loading hot states via block replay doesn't care about the schema version, so we can use
// the DB's current method for this.
let update_cache = false;
db.load_hot_state_using_replay(
base_state,
summary.slot,
summary.latest_block_root,
StatePayloadStatus::Pending,
update_cache,
)
.map(Some)
}
pub fn get_full_state_v22<KV: KeyValueStore<E>, E: EthSpec>(
db: &KV,
state_root: &Hash256,
spec: &ChainSpec,
) -> Result<Option<BeaconState<E>>, Error> {
match db.get_bytes(DBColumn::BeaconState, state_root.as_slice())? {
Some(bytes) => {
let container = StorageContainer::from_ssz_bytes(&bytes, spec)?;
Ok(Some(container.try_into()?))
}
None => Ok(None),
}
}
/// A container for storing `BeaconState` components.
///
/// DEPRECATED.
#[derive(Encode)]
pub struct StorageContainer<E: EthSpec> {
state: BeaconState<E>,
committee_caches: Vec<Arc<CommitteeCache>>,
}
impl<E: EthSpec> StorageContainer<E> {
/// Create a new instance for storing a `BeaconState`.
pub fn new(state: &BeaconState<E>) -> Self {
Self {
state: state.clone(),
committee_caches: state.committee_caches().to_vec(),
}
}
pub fn from_ssz_bytes(bytes: &[u8], spec: &ChainSpec) -> Result<Self, ssz::DecodeError> {
// We need to use the slot-switching `from_ssz_bytes` of `BeaconState`, which doesn't
// compose with the other SSZ utils, so we duplicate some parts of `ssz_derive` here.
let mut builder = ssz::SszDecoderBuilder::new(bytes);
builder.register_anonymous_variable_length_item()?;
builder.register_type::<Vec<CommitteeCache>>()?;
let mut decoder = builder.build()?;
let state = decoder.decode_next_with(|bytes| BeaconState::from_ssz_bytes(bytes, spec))?;
let committee_caches = decoder.decode_next()?;
Ok(Self {
state,
committee_caches,
})
}
}
impl<E: EthSpec> TryInto<BeaconState<E>> for StorageContainer<E> {
type Error = Error;
fn try_into(mut self) -> Result<BeaconState<E>, Error> {
let mut state = self.state;
for i in (0..CACHED_EPOCHS).rev() {
if i >= self.committee_caches.len() {
return Err(Error::SszDecodeError(DecodeError::BytesInvalid(
"Insufficient committees for BeaconState".to_string(),
)));
};
state.committee_caches_mut()[i] = self.committee_caches.remove(i);
}
Ok(state)
}
}
/// The checkpoint used for pruning the database.
///
/// Updated whenever pruning is successful.
#[derive(Debug, Clone, Copy, PartialEq, Eq)]
pub struct PruningCheckpoint {
pub checkpoint: Checkpoint,
}
impl StoreItem for PruningCheckpoint {
fn db_column() -> DBColumn {
DBColumn::BeaconMeta
}
fn as_store_bytes(&self) -> Vec<u8> {
self.checkpoint.as_ssz_bytes()
}
fn from_store_bytes(bytes: &[u8]) -> Result<Self, Error> {
Ok(PruningCheckpoint {
checkpoint: Checkpoint::from_ssz_bytes(bytes)?,
})
}
}
pub fn upgrade_to_v24<T: BeaconChainTypes>(
db: Arc<HotColdDB<T::EthSpec, T::HotStore, T::ColdStore>>,
) -> Result<Vec<KeyValueStoreOp>, Error> {
let mut migrate_ops = vec![];
let split = db.get_split_info();
let hot_hdiff_start_slot = split.slot;
// Delete the `PruningCheckpoint` (no longer used).
migrate_ops.push(KeyValueStoreOp::DeleteKey(
DBColumn::BeaconMeta,
PRUNING_CHECKPOINT_KEY.as_slice().to_vec(),
));
// Sanity check to make sure the HDiff grid is aligned with the epoch start
if hot_hdiff_start_slot % T::EthSpec::slots_per_epoch() != 0 {
return Err(Error::MigrationError(format!(
"hot_hdiff_start_slot is not first slot in epoch {hot_hdiff_start_slot}"
)));
}
// After V24 hot tree states, the in-memory `anchor_info.anchor_slot` is the start slot of the
// hot HDiff grid. Before the migration, it's set to the slot of the anchor state in the DB:
// - the genesis state on a genesis sync, or
// - the checkpoint state on a checkpoint sync.
//
// If the node has been running for a while the `anchor_slot` might be less than the finalized
// checkpoint. This upgrade constructs a grid only with unfinalized states, rooted in the
// current finalized state. So we set the `anchor_slot` to `split.slot` to root the grid in the
// current finalized state. Each migration sets the split to
// ```
// Split { slot: finalized_state.slot(), state_root: finalized_state_root }
// ```
{
let anchor_info = db.get_anchor_info();
// If the node is already an archive node, we can set the anchor slot to 0 and copy
// snapshots and diffs from the freezer DB to the hot DB in order to establish an initial
// hot grid that is aligned/"perfect" (no `start_slot`/`anchor_slot` to worry about).
//
// This only works if all of the following are true:
//
// - We have the previous snapshot for the split state stored in the freezer DB, i.e.
// if `previous_snapshot_slot >= state_upper_limit`.
// - The split state itself will be stored as a diff or snapshot in the new grid. We choose
// not to support a split state that requires block replay, because computing its previous
// state root from the DAG is not straight-forward.
let dummy_start_slot = Slot::new(0);
let closest_layer_points = db
.hierarchy
.closest_layer_points(split.slot, dummy_start_slot);
let previous_snapshot_slot =
closest_layer_points
.iter()
.copied()
.min()
.ok_or(Error::MigrationError(
"closest_layer_points must not be empty".to_string(),
))?;
if previous_snapshot_slot >= anchor_info.state_upper_limit
&& db
.hierarchy
.storage_strategy(split.slot, dummy_start_slot)
.is_ok_and(|strategy| !strategy.is_replay_from())
{
info!(
%previous_snapshot_slot,
split_slot = %split.slot,
"Aligning hot diff grid to freezer"
);
// Set anchor slot to 0 in case it was set to something else by a previous checkpoint
// sync.
let mut new_anchor_info = anchor_info.clone();
new_anchor_info.anchor_slot = Slot::new(0);
// Update the anchor on disk atomically if migration is successful
migrate_ops.push(db.compare_and_set_anchor_info(anchor_info, new_anchor_info)?);
// Copy each of the freezer layers to the hot DB in slot ascending order.
for layer_slot in closest_layer_points.into_iter().rev() {
// Do not try to load the split state itself from the freezer, it won't be there.
// It will be migrated in the main loop below.
if layer_slot == split.slot {
continue;
}
let mut freezer_state = db.load_cold_state_by_slot(layer_slot)?;
let state_root = freezer_state.canonical_root()?;
let mut state_ops = vec![];
db.store_hot_state(&state_root, &freezer_state, &mut state_ops)?;
db.hot_db.do_atomically(state_ops)?;
}
} else {
// Otherwise for non-archive nodes, set the anchor slot for the hot grid to the current
// split slot (the oldest slot available).
let mut new_anchor_info = anchor_info.clone();
new_anchor_info.anchor_slot = hot_hdiff_start_slot;
// Update the anchor in disk atomically if migration is successful
migrate_ops.push(db.compare_and_set_anchor_info(anchor_info, new_anchor_info)?);
}
}
let state_summaries_dag = new_dag::<T>(&db)?;
// We compute the state summaries DAG outside of a DB migration. Therefore if the DB is properly
// prunned, it should have a single root equal to the split.
let state_summaries_dag_roots = state_summaries_dag.tree_roots();
if state_summaries_dag_roots.len() == 1 {
let (root_summary_state_root, root_summary) =
state_summaries_dag_roots.first().expect("len == 1");
if *root_summary_state_root != split.state_root {
warn!(
?root_summary_state_root,
?root_summary,
?split,
"State summaries DAG root is not the split"
);
}
} else {
warn!(
location = "migration",
state_summaries_dag_roots = ?state_summaries_dag_roots,
"State summaries DAG found more than one root"
);
}
// Sort summaries by slot so we have their ancestor diffs already stored when we store them.
// If the summaries are sorted topologically we can insert them into the DB like if they were a
// new state, re-using existing code. As states are likely to be sequential the diff cache
// should kick in making the migration more efficient. If we just iterate the column of
// summaries we may get distance state of each iteration.
let summaries_by_slot = state_summaries_dag.summaries_by_slot_ascending();
debug!(
summaries_count = state_summaries_dag.summaries_count(),
slots_count = summaries_by_slot.len(),
min_slot = ?summaries_by_slot.first_key_value().map(|(slot, _)| slot),
max_slot = ?summaries_by_slot.last_key_value().map(|(slot, _)| slot),
?state_summaries_dag_roots,
%hot_hdiff_start_slot,
split_state_root = ?split.state_root,
"Starting hot states migration"
);
// Upgrade all hot DB state summaries to the new type:
// - Set all summaries of boundary states to `Snapshot` type
// - Set all others to `Replay` pointing to `epoch_boundary_state_root`
let mut diffs_written = 0;
let mut summaries_written = 0;
let mut last_log_time = Instant::now();
for (slot, old_hot_state_summaries) in summaries_by_slot {
for (state_root, old_summary) in old_hot_state_summaries {
if slot < hot_hdiff_start_slot {
// To reach here, there must be some pruning issue with the DB where we still have
// hot states below the split slot. This states can't be migrated as we can't compute
// a storage strategy for them. After this if else block, the summary and state are
// scheduled for deletion.
debug!(
%slot,
?state_root,
"Ignoring state summary prior to split slot"
);
} else {
// 1. Store snapshot or diff at this slot (if required).
let storage_strategy = db.hot_storage_strategy(slot)?;
debug!(
%slot,
?state_root,
?storage_strategy,
"Migrating state summary"
);
match storage_strategy {
StorageStrategy::DiffFrom(_) | StorageStrategy::Snapshot => {
// Load the state and re-store it as a snapshot or diff.
let state = get_state_v22::<T>(&db, &state_root, &db.spec)?
.ok_or(Error::MissingState(state_root))?;
// Store immediately so that future diffs can load and diff from it.
let mut ops = vec![];
// We must commit the hot state summary immediately, otherwise we can't diff
// against it and future writes will fail. That's why we write the new hot
// summaries in a different column to have both new and old data present at
// once. Otherwise if the process crashes during the migration the database will
// be broken.
db.store_hot_state_summary(&state_root, &state, &mut ops)?;
db.store_hot_state_diffs(&state_root, &state, &mut ops)?;
db.hot_db.do_atomically(ops)?;
diffs_written += 1;
}
StorageStrategy::ReplayFrom(diff_base_slot) => {
// Optimization: instead of having to load the state of each summary we load x32
// less states by manually computing the HotStateSummary roots using the
// computed state dag.
//
// No need to store diffs for states that will be reconstructed by replaying
// blocks.
//
// 2. Convert the summary to the new format.
if state_root == split.state_root {
return Err(Error::MigrationError(
"unreachable: split state should be stored as a snapshot or diff"
.to_string(),
));
}
let previous_state_root = state_summaries_dag
.previous_state_root(state_root)
.map_err(|e| {
Error::MigrationError(format!(
"error computing previous_state_root {e:?}"
))
})?;
let diff_base_state = OptionalDiffBaseState::new(
diff_base_slot,
state_summaries_dag
.ancestor_state_root_at_slot(state_root, diff_base_slot)
.map_err(|e| {
Error::MigrationError(format!(
"error computing ancestor_state_root_at_slot \
({state_root:?}, {diff_base_slot}): {e:?}"
))
})?,
);
let new_summary = HotStateSummary {
slot,
latest_block_root: old_summary.latest_block_root,
latest_block_slot: old_summary.latest_block_slot,
previous_state_root,
diff_base_state,
};
let op = new_summary.as_kv_store_op(state_root);
// It's not necessary to immediately commit the summaries of states that are
// ReplayFrom. However we do so for simplicity.
db.hot_db.do_atomically(vec![op])?;
}
}
}
// 3. Stage old data for deletion.
if slot % T::EthSpec::slots_per_epoch() == 0 {
migrate_ops.push(KeyValueStoreOp::DeleteKey(
DBColumn::BeaconState,
state_root.as_slice().to_vec(),
));
}
// Delete previous summaries
migrate_ops.push(KeyValueStoreOp::DeleteKey(
DBColumn::BeaconStateSummary,
state_root.as_slice().to_vec(),
));
summaries_written += 1;
if last_log_time.elapsed() > Duration::from_secs(5) {
last_log_time = Instant::now();
info!(
diffs_written,
summaries_written,
summaries_count = state_summaries_dag.summaries_count(),
"Hot states migration in progress"
);
}
}
}
info!(
diffs_written,
summaries_written,
summaries_count = state_summaries_dag.summaries_count(),
"Hot states migration complete"
);
Ok(migrate_ops)
}
pub fn downgrade_from_v24<T: BeaconChainTypes>(
db: Arc<HotColdDB<T::EthSpec, T::HotStore, T::ColdStore>>,
) -> Result<Vec<KeyValueStoreOp>, Error> {
let state_summaries = db
.load_hot_state_summaries()?
.into_iter()
.map(|(state_root, summary)| (state_root, summary.into()))
.collect::<Vec<(Hash256, DAGStateSummary)>>();
info!(
summaries_count = state_summaries.len(),
"DB downgrade of v24 state summaries started"
);
let state_summaries_dag = StateSummariesDAG::new(state_summaries)
.map_err(|e| Error::MigrationError(format!("Error on new StateSumariesDAG {e:?}")))?;
let mut migrate_ops = vec![];
let mut states_written = 0;
let mut summaries_written = 0;
let mut summaries_skipped = 0;
let mut last_log_time = Instant::now();
// Rebuild the PruningCheckpoint from the split.
let split = db.get_split_info();
let pruning_checkpoint = PruningCheckpoint {
checkpoint: Checkpoint {
epoch: split.slot.epoch(T::EthSpec::slots_per_epoch()),
root: split.block_root,
},
};
migrate_ops.push(pruning_checkpoint.as_kv_store_op(PRUNING_CHECKPOINT_KEY));
// Convert state summaries back to the old format.
for (state_root, summary) in state_summaries_dag
.summaries_by_slot_ascending()
.into_iter()
.flat_map(|(_, summaries)| summaries)
{
// No need to migrate any states prior to the split. The v22 schema does not need them, and
// they would generate warnings about a disjoint DAG when re-upgrading to V24.
if summary.slot < split.slot {
debug!(
slot = %summary.slot,
?state_root,
"Skipping migration of pre-split state"
);
summaries_skipped += 1;
continue;
}
// If boundary state: persist.
// Do not cache these states as they are unlikely to be relevant later.
let update_cache = false;
if summary.slot % T::EthSpec::slots_per_epoch() == 0 {
let (state, _) = db
.load_hot_state(&state_root, update_cache)?
.ok_or(Error::MissingState(state_root))?;
// Immediately commit the state, so we don't OOM. It's stored in a different
// column so if the migration crashes we'll just store extra harmless junk in the DB.
let mut state_write_ops = vec![];
store_full_state_v22(&state_root, &state, &mut state_write_ops)?;
db.hot_db.do_atomically(state_write_ops)?;
states_written += 1;
}
// Persist old summary.
let epoch_boundary_state_slot = summary.slot - summary.slot % T::EthSpec::slots_per_epoch();
let old_summary = HotStateSummaryV22 {
slot: summary.slot,
latest_block_root: summary.latest_block_root,
epoch_boundary_state_root: state_summaries_dag
.ancestor_state_root_at_slot(state_root, epoch_boundary_state_slot)
.map_err(|e| {
Error::MigrationError(format!(
"error computing ancestor_state_root_at_slot({state_root:?}, {epoch_boundary_state_slot}) {e:?}"
))
})?,
};
migrate_ops.push(KeyValueStoreOp::PutKeyValue(
DBColumn::BeaconStateSummary,
state_root.as_slice().to_vec(),
old_summary.as_ssz_bytes(),
));
summaries_written += 1;
if last_log_time.elapsed() > Duration::from_secs(5) {
last_log_time = Instant::now();
info!(
states_written,
summaries_written,
summaries_count = state_summaries_dag.summaries_count(),
"DB downgrade of v24 state summaries in progress"
);
}
}
// Delete all V24 schema data. We do this outside the loop over summaries to ensure we cover
// every piece of data and to simplify logic around skipping certain summaries that do not get
// migrated.
for db_column in [
DBColumn::BeaconStateHotSummary,
DBColumn::BeaconStateHotDiff,
DBColumn::BeaconStateHotSnapshot,
] {
for key in db.hot_db.iter_column_keys::<Hash256>(db_column) {
let state_root = key?;
migrate_ops.push(KeyValueStoreOp::DeleteKey(
db_column,
state_root.as_slice().to_vec(),
));
}
}
info!(
states_written,
summaries_written,
summaries_skipped,
summaries_count = state_summaries_dag.summaries_count(),
"DB downgrade of v24 state summaries completed"
);
Ok(migrate_ops)
}
fn new_dag<T: BeaconChainTypes>(
db: &HotColdDB<T::EthSpec, T::HotStore, T::ColdStore>,
) -> Result<StateSummariesDAG, Error> {
// Collect all sumaries for unfinalized states
let state_summaries_v22 = db
.hot_db
// Collect summaries from the legacy V22 column BeaconStateSummary
.iter_column::<Hash256>(DBColumn::BeaconStateSummary)
.map(|res| {
let (key, value) = res?;
let state_root: Hash256 = key;
let summary = HotStateSummaryV22::from_ssz_bytes(&value)?;
let block_root = summary.latest_block_root;
// Read blocks to get the block slot and parent root. In Holesky forced finalization it
// took 5100 ms to read 15072 state summaries, so it's not really necessary to
// de-duplicate block reads.
let block = db
.get_blinded_block(&block_root)?
.ok_or(Error::MissingBlock(block_root))?;
Ok((
state_root,
DAGStateSummaryV22 {
slot: summary.slot,
latest_block_root: summary.latest_block_root,
block_slot: block.slot(),
block_parent_root: block.parent_root(),
},
))
})
.collect::<Result<Vec<_>, Error>>()?;
StateSummariesDAG::new_from_v22(state_summaries_v22)
.map_err(|e| Error::MigrationError(format!("error computing states summaries dag {e:?}")))
}

View File

@@ -1,20 +0,0 @@
use store::{DBColumn, Error, KeyValueStoreOp};
use tracing::info;
use types::Hash256;
pub const ETH1_CACHE_DB_KEY: Hash256 = Hash256::ZERO;
/// Delete the on-disk eth1 data.
pub fn upgrade_to_v25() -> Result<Vec<KeyValueStoreOp>, Error> {
info!("Deleting eth1 data from disk for v25 DB upgrade");
Ok(vec![KeyValueStoreOp::DeleteKey(
DBColumn::Eth1Cache,
ETH1_CACHE_DB_KEY.as_slice().to_vec(),
)])
}
/// No-op: we don't need to recreate on-disk eth1 data, as previous versions gracefully handle
/// data missing from disk.
pub fn downgrade_from_v25() -> Result<Vec<KeyValueStoreOp>, Error> {
Ok(vec![])
}

View File

@@ -1,91 +0,0 @@
use crate::BeaconChainTypes;
use crate::custody_context::CustodyContextSsz;
use crate::persisted_custody::{CUSTODY_DB_KEY, PersistedCustody};
use ssz::{Decode, Encode};
use ssz_derive::{Decode, Encode};
use std::sync::Arc;
use store::{DBColumn, Error, HotColdDB, KeyValueStoreOp, StoreItem};
use tracing::info;
#[derive(Debug, Encode, Decode, Clone)]
pub(crate) struct CustodyContextSszV24 {
pub(crate) validator_custody_at_head: u64,
pub(crate) persisted_is_supernode: bool,
}
pub(crate) struct PersistedCustodyV24(CustodyContextSszV24);
impl StoreItem for PersistedCustodyV24 {
fn db_column() -> DBColumn {
DBColumn::CustodyContext
}
fn as_store_bytes(&self) -> Vec<u8> {
self.0.as_ssz_bytes()
}
fn from_store_bytes(bytes: &[u8]) -> Result<Self, Error> {
let custody_context = CustodyContextSszV24::from_ssz_bytes(bytes)?;
Ok(PersistedCustodyV24(custody_context))
}
}
/// Upgrade the `CustodyContext` entry to v26.
pub fn upgrade_to_v26<T: BeaconChainTypes>(
db: Arc<HotColdDB<T::EthSpec, T::HotStore, T::ColdStore>>,
) -> Result<Vec<KeyValueStoreOp>, Error> {
let ops = if db.spec.is_peer_das_scheduled() {
match db.get_item::<PersistedCustodyV24>(&CUSTODY_DB_KEY) {
Ok(Some(PersistedCustodyV24(ssz_v24))) => {
info!("Migrating `CustodyContext` to v26 schema");
let custody_context_v2 = CustodyContextSsz {
validator_custody_at_head: ssz_v24.validator_custody_at_head,
persisted_is_supernode: ssz_v24.persisted_is_supernode,
epoch_validator_custody_requirements: vec![],
};
vec![KeyValueStoreOp::PutKeyValue(
DBColumn::CustodyContext,
CUSTODY_DB_KEY.as_slice().to_vec(),
PersistedCustody(custody_context_v2).as_store_bytes(),
)]
}
_ => {
vec![]
}
}
} else {
// Delete it from db if PeerDAS hasn't been scheduled
vec![KeyValueStoreOp::DeleteKey(
DBColumn::CustodyContext,
CUSTODY_DB_KEY.as_slice().to_vec(),
)]
};
Ok(ops)
}
pub fn downgrade_from_v26<T: BeaconChainTypes>(
db: Arc<HotColdDB<T::EthSpec, T::HotStore, T::ColdStore>>,
) -> Result<Vec<KeyValueStoreOp>, Error> {
let res = db.get_item::<PersistedCustody>(&CUSTODY_DB_KEY);
let ops = match res {
Ok(Some(PersistedCustody(ssz_v26))) => {
info!("Migrating `CustodyContext` back from v26 schema");
let custody_context_v24 = CustodyContextSszV24 {
validator_custody_at_head: ssz_v26.validator_custody_at_head,
persisted_is_supernode: ssz_v26.persisted_is_supernode,
};
vec![KeyValueStoreOp::PutKeyValue(
DBColumn::CustodyContext,
CUSTODY_DB_KEY.as_slice().to_vec(),
PersistedCustodyV24(custody_context_v24).as_store_bytes(),
)]
}
_ => {
// no op if it's not on the db, as previous versions gracefully handle data missing from disk.
vec![]
}
};
Ok(ops)
}

View File

@@ -1,26 +0,0 @@
use crate::BeaconChainTypes;
use std::sync::Arc;
use store::{Error, HotColdDB, metadata::SchemaVersion};
/// Add `DataColumnCustodyInfo` entry to v27.
pub fn upgrade_to_v27<T: BeaconChainTypes>(
db: Arc<HotColdDB<T::EthSpec, T::HotStore, T::ColdStore>>,
) -> Result<(), Error> {
if db.spec.is_peer_das_scheduled() {
db.put_data_column_custody_info(None)?;
db.store_schema_version_atomically(SchemaVersion(27), vec![])?;
}
Ok(())
}
pub fn downgrade_from_v27<T: BeaconChainTypes>(
db: Arc<HotColdDB<T::EthSpec, T::HotStore, T::ColdStore>>,
) -> Result<(), Error> {
if db.spec.is_peer_das_scheduled() {
return Err(Error::MigrationError(
"Cannot downgrade from v27 if peerDAS is scheduled".to_string(),
));
}
Ok(())
}

View File

@@ -1,152 +0,0 @@
use crate::{
BeaconChain, BeaconChainTypes, BeaconForkChoiceStore, PersistedForkChoiceStoreV17,
beacon_chain::FORK_CHOICE_DB_KEY,
persisted_fork_choice::{PersistedForkChoiceV17, PersistedForkChoiceV28},
summaries_dag::{DAGStateSummary, StateSummariesDAG},
};
use fork_choice::{ForkChoice, ForkChoiceStore, ResetPayloadStatuses};
use std::sync::Arc;
use store::{Error, HotColdDB, KeyValueStoreOp, StoreItem};
use tracing::{info, warn};
use types::{EthSpec, Hash256};
/// Upgrade `PersistedForkChoice` from V17 to V28.
pub fn upgrade_to_v28<T: BeaconChainTypes>(
db: Arc<HotColdDB<T::EthSpec, T::HotStore, T::ColdStore>>,
) -> Result<Vec<KeyValueStoreOp>, Error> {
let Some(persisted_fork_choice_v17) =
db.get_item::<PersistedForkChoiceV17>(&FORK_CHOICE_DB_KEY)?
else {
warn!("No fork choice found to upgrade to v28");
return Ok(vec![]);
};
// Load state DAG in order to compute justified checkpoint roots.
let state_summaries_dag = {
let state_summaries = db
.load_hot_state_summaries()?
.into_iter()
.map(|(state_root, summary)| (state_root, summary.into()))
.collect::<Vec<(Hash256, DAGStateSummary)>>();
StateSummariesDAG::new(state_summaries).map_err(|e| {
Error::MigrationError(format!("Error loading state summaries DAG: {e:?}"))
})?
};
// Determine the justified state roots.
let justified_checkpoint = persisted_fork_choice_v17
.fork_choice_store_v17
.justified_checkpoint;
let justified_block_root = justified_checkpoint.root;
let justified_slot = justified_checkpoint
.epoch
.start_slot(T::EthSpec::slots_per_epoch());
let justified_state_root = state_summaries_dag
.state_root_at_slot(justified_block_root, justified_slot)
.ok_or_else(|| {
Error::MigrationError(format!(
"Missing state root for justified slot {justified_slot} with latest_block_root \
{justified_block_root:?}"
))
})?;
let unrealized_justified_checkpoint = persisted_fork_choice_v17
.fork_choice_store_v17
.unrealized_justified_checkpoint;
let unrealized_justified_block_root = unrealized_justified_checkpoint.root;
let unrealized_justified_slot = unrealized_justified_checkpoint
.epoch
.start_slot(T::EthSpec::slots_per_epoch());
let unrealized_justified_state_root = state_summaries_dag
.state_root_at_slot(unrealized_justified_block_root, unrealized_justified_slot)
.ok_or_else(|| {
Error::MigrationError(format!(
"Missing state root for unrealized justified slot {unrealized_justified_slot} \
with latest_block_root {unrealized_justified_block_root:?}"
))
})?;
let fc_store = BeaconForkChoiceStore::from_persisted_v17(
persisted_fork_choice_v17.fork_choice_store_v17,
justified_state_root,
unrealized_justified_state_root,
db.clone(),
)
.map_err(|e| {
Error::MigrationError(format!(
"Error loading fork choice store from persisted: {e:?}"
))
})?;
info!(
?justified_state_root,
%justified_slot,
"Added justified state root to fork choice"
);
// Construct top-level ForkChoice struct using the patched fork choice store, and the converted
// proto array.
let reset_payload_statuses = ResetPayloadStatuses::OnlyWithInvalidPayload;
let fork_choice = ForkChoice::from_persisted(
persisted_fork_choice_v17.fork_choice_v17.try_into()?,
reset_payload_statuses,
fc_store,
db.get_chain_spec(),
)
.map_err(|e| Error::MigrationError(format!("Unable to build ForkChoice: {e:?}")))?;
let ops = vec![BeaconChain::<T>::persist_fork_choice_in_batch_standalone(
&fork_choice,
db.get_config(),
)?];
info!("Upgraded fork choice for DB schema v28");
Ok(ops)
}
pub fn downgrade_from_v28<T: BeaconChainTypes>(
db: Arc<HotColdDB<T::EthSpec, T::HotStore, T::ColdStore>>,
) -> Result<Vec<KeyValueStoreOp>, Error> {
let reset_payload_statuses = ResetPayloadStatuses::OnlyWithInvalidPayload;
let Some(fork_choice) =
BeaconChain::<T>::load_fork_choice(db.clone(), reset_payload_statuses, db.get_chain_spec())
.map_err(|e| Error::MigrationError(format!("Unable to load fork choice: {e:?}")))?
else {
warn!("No fork choice to downgrade");
return Ok(vec![]);
};
// Recreate V28 persisted fork choice, then convert each field back to its V17 version.
let persisted_fork_choice = PersistedForkChoiceV28 {
fork_choice: fork_choice.to_persisted(),
fork_choice_store: fork_choice.fc_store().to_persisted(),
};
let justified_balances = fork_choice.fc_store().justified_balances();
// 1. Create `proto_array::PersistedForkChoiceV17`.
let fork_choice_v17: fork_choice::PersistedForkChoiceV17 = (
persisted_fork_choice.fork_choice,
justified_balances.clone(),
)
.into();
let fork_choice_store_v17: PersistedForkChoiceStoreV17 = (
persisted_fork_choice.fork_choice_store,
justified_balances.clone(),
)
.into();
let persisted_fork_choice_v17 = PersistedForkChoiceV17 {
fork_choice_v17,
fork_choice_store_v17,
};
let ops = vec![persisted_fork_choice_v17.as_kv_store_op(FORK_CHOICE_DB_KEY)];
info!("Downgraded fork choice for DB schema v28");
Ok(ops)
}

View File

@@ -14,14 +14,6 @@ pub struct DAGStateSummary {
pub previous_state_root: Hash256,
}
#[derive(Debug, Clone, Copy)]
pub struct DAGStateSummaryV22 {
pub slot: Slot,
pub latest_block_root: Hash256,
pub block_slot: Slot,
pub block_parent_root: Hash256,
}
pub struct StateSummariesDAG {
// state_root -> state_summary
state_summaries_by_state_root: HashMap<Hash256, DAGStateSummary>,
@@ -40,10 +32,6 @@ pub enum Error {
new_state_summary: (Slot, Hash256),
},
MissingStateSummary(Hash256),
MissingStateSummaryByBlockRoot {
state_root: Hash256,
latest_block_root: Hash256,
},
MissingChildStateRoot(Hash256),
RequestedSlotAboveSummary {
starting_state_root: Hash256,
@@ -109,89 +97,6 @@ impl StateSummariesDAG {
})
}
/// Computes a DAG from a sequence of state summaries, including their parent block
/// relationships.
///
/// - Expects summaries to be contiguous per slot: there must exist a summary at every slot
/// of each tree branch
/// - Maybe include multiple disjoint trees. The root of each tree will have a ZERO parent state
/// root, which will error later when calling `previous_state_root`.
pub fn new_from_v22(
state_summaries_v22: Vec<(Hash256, DAGStateSummaryV22)>,
) -> Result<Self, Error> {
// Group them by latest block root, and sorted state slot
let mut state_summaries_by_block_root = HashMap::<_, BTreeMap<_, _>>::new();
for (state_root, summary) in state_summaries_v22.iter() {
let summaries = state_summaries_by_block_root
.entry(summary.latest_block_root)
.or_default();
// Sanity check to ensure no duplicate summaries for the tuple (block_root, state_slot)
match summaries.entry(summary.slot) {
Entry::Vacant(entry) => {
entry.insert((state_root, summary));
}
Entry::Occupied(existing) => {
return Err(Error::DuplicateStateSummary {
block_root: summary.latest_block_root,
existing_state_summary: (summary.slot, *state_root).into(),
new_state_summary: (*existing.key(), *existing.get().0),
});
}
}
}
let state_summaries = state_summaries_v22
.iter()
.map(|(state_root, summary)| {
let previous_state_root = if summary.slot == 0 {
Hash256::ZERO
} else {
let previous_slot = summary.slot - 1;
// Check the set of states in the same state's block root
let same_block_root_summaries = state_summaries_by_block_root
.get(&summary.latest_block_root)
// Should never error: we construct the HashMap here and must have at least
// one entry per block root
.ok_or(Error::MissingStateSummaryByBlockRoot {
state_root: *state_root,
latest_block_root: summary.latest_block_root,
})?;
if let Some((state_root, _)) = same_block_root_summaries.get(&previous_slot) {
// Skipped slot: block root at previous slot is the same as latest block root.
**state_root
} else {
// Common case: not a skipped slot.
//
// If we can't find a state summmary for the parent block and previous slot,
// then there is some amount of disjointedness in the DAG. We set the parent
// state root to 0x0 in this case, and will prune any dangling states.
let parent_block_root = summary.block_parent_root;
state_summaries_by_block_root
.get(&parent_block_root)
.and_then(|parent_block_summaries| {
parent_block_summaries.get(&previous_slot)
})
.map_or(Hash256::ZERO, |(parent_state_root, _)| **parent_state_root)
}
};
Ok((
*state_root,
DAGStateSummary {
slot: summary.slot,
latest_block_root: summary.latest_block_root,
latest_block_slot: summary.block_slot,
previous_state_root,
},
))
})
.collect::<Result<Vec<_>, _>>()?;
Self::new(state_summaries)
}
// Returns all non-unique latest block roots of a given set of states
pub fn blocks_of_states<'a, I: Iterator<Item = &'a Hash256>>(
&self,
@@ -379,106 +284,3 @@ impl From<HotStateSummary> for DAGStateSummary {
}
}
}
#[cfg(test)]
mod tests {
use super::{DAGStateSummaryV22, Error, StateSummariesDAG};
use bls::FixedBytesExtended;
use types::{Hash256, Slot};
fn root(n: u64) -> Hash256 {
Hash256::from_low_u64_le(n)
}
#[test]
fn new_from_v22_empty() {
StateSummariesDAG::new_from_v22(vec![]).unwrap();
}
fn assert_previous_state_root_is_zero(dag: &StateSummariesDAG, root: Hash256) {
assert!(matches!(
dag.previous_state_root(root).unwrap_err(),
Error::RootUnknownPreviousStateRoot { .. }
));
}
#[test]
fn new_from_v22_one_state() {
let root_a = root(0xa);
let root_1 = root(1);
let root_2 = root(2);
let summary_1 = DAGStateSummaryV22 {
slot: Slot::new(1),
latest_block_root: root_1,
block_parent_root: root_2,
block_slot: Slot::new(1),
};
let dag = StateSummariesDAG::new_from_v22(vec![(root_a, summary_1)]).unwrap();
// The parent of the root summary is ZERO
assert_previous_state_root_is_zero(&dag, root_a);
}
#[test]
fn new_from_v22_multiple_states() {
let dag = StateSummariesDAG::new_from_v22(vec![
(
root(0xa),
DAGStateSummaryV22 {
slot: Slot::new(3),
latest_block_root: root(3),
block_parent_root: root(1),
block_slot: Slot::new(3),
},
),
(
root(0xb),
DAGStateSummaryV22 {
slot: Slot::new(4),
latest_block_root: root(4),
block_parent_root: root(3),
block_slot: Slot::new(4),
},
),
// fork 1
(
root(0xc),
DAGStateSummaryV22 {
slot: Slot::new(5),
latest_block_root: root(5),
block_parent_root: root(4),
block_slot: Slot::new(5),
},
),
// fork 2
// skipped slot
(
root(0xd),
DAGStateSummaryV22 {
slot: Slot::new(5),
latest_block_root: root(4),
block_parent_root: root(3),
block_slot: Slot::new(4),
},
),
// normal slot
(
root(0xe),
DAGStateSummaryV22 {
slot: Slot::new(6),
latest_block_root: root(6),
block_parent_root: root(4),
block_slot: Slot::new(6),
},
),
])
.unwrap();
// The parent of the root summary is ZERO
assert_previous_state_root_is_zero(&dag, root(0xa));
assert_eq!(dag.previous_state_root(root(0xc)).unwrap(), root(0xb));
assert_eq!(dag.previous_state_root(root(0xd)).unwrap(), root(0xb));
assert_eq!(dag.previous_state_root(root(0xe)).unwrap(), root(0xd));
}
}