cerc-io/plugeth
12
0
Fork 1
Code Issues 4 Pull Requests Packages Projects Releases Wiki Activity

core, eth, trie: filter out boundary nodes and remove dangling nodes in stacktrie (#28327)

Browse Source 
* core, eth, trie: filter out boundary nodes in stacktrie

* eth/protocol/snap: add comments

* Update trie/stacktrie.go

Co-authored-by: Martin Holst Swende <martin@swende.se>

* eth, trie: remove onBoundary callback

* eth/protocols/snap: keep complete boundary nodes

* eth/protocols/snap: skip healing if the storage trie is already complete

* eth, trie: add more metrics

* eth, trie: address comment

---------

Co-authored-by: Martin Holst Swende <martin@swende.se>
This commit is contained in:
rjl493456442
2023-10-23 18:31:56 +03:00
committed by GitHub
co-authored by Martin Holst Swende
parent 43e6a3c196
commit ab04aeb855
5 changed files with 312 additions and 11 deletions
Download Patch File Download Diff File Expand all files Collapse all files
eth/protocols/snap/metrics.go
+28
View File
@@ -26,4 +26,32 @@ var (
IngressRegistrationErrorMeter = metrics.NewRegisteredMeter(ingressRegistrationErrorName, nil)
EgressRegistrationErrorMeter = metrics.NewRegisteredMeter(egressRegistrationErrorName, nil)
// deletionGauge is the metric to track how many trie node deletions
// are performed in total during the sync process.
deletionGauge = metrics.NewRegisteredGauge("eth/protocols/snap/sync/delete", nil)
// lookupGauge is the metric to track how many trie node lookups are
// performed to determine if node needs to be deleted.
lookupGauge = metrics.NewRegisteredGauge("eth/protocols/snap/sync/lookup", nil)
// boundaryAccountNodesGauge is the metric to track how many boundary trie
// nodes in account trie are met.
boundaryAccountNodesGauge = metrics.NewRegisteredGauge("eth/protocols/snap/sync/boundary/account", nil)
// boundaryAccountNodesGauge is the metric to track how many boundary trie
// nodes in storage tries are met.
boundaryStorageNodesGauge = metrics.NewRegisteredGauge("eth/protocols/snap/sync/boundary/storage", nil)
// smallStorageGauge is the metric to track how many storages are small enough
// to retrieved in one or two request.
smallStorageGauge = metrics.NewRegisteredGauge("eth/protocols/snap/sync/storage/small", nil)
// largeStorageGauge is the metric to track how many storages are large enough
// to retrieved concurrently.
largeStorageGauge = metrics.NewRegisteredGauge("eth/protocols/snap/sync/storage/large", nil)
// skipStorageHealingGauge is the metric to track how many storages are retrieved
// in multiple requests but healing is not necessary.
skipStorageHealingGauge = metrics.NewRegisteredGauge("eth/protocols/snap/sync/storage/noheal", nil)
)
eth/protocols/snap/sync.go
+94 -5
View File
@@ -716,6 +716,19 @@ func (s *Syncer) Sync(root common.Hash, cancel chan struct{}) error {
}
}
// cleanPath is used to remove the dangling nodes in the stackTrie.
func (s *Syncer) cleanPath(batch ethdb.Batch, owner common.Hash, path []byte) {
if owner == (common.Hash{}) && rawdb.ExistsAccountTrieNode(s.db, path) {
rawdb.DeleteAccountTrieNode(batch, path)
deletionGauge.Inc(1)
}
if owner != (common.Hash{}) && rawdb.ExistsStorageTrieNode(s.db, owner, path) {
rawdb.DeleteStorageTrieNode(batch, owner, path)
deletionGauge.Inc(1)
}
lookupGauge.Inc(1)
}
// loadSyncStatus retrieves a previously aborted sync status from the database,
// or generates a fresh one if none is available.
func (s *Syncer) loadSyncStatus() {
@@ -742,6 +755,17 @@ func (s *Syncer) loadSyncStatus() {
options = options.WithWriter(func(path []byte, hash common.Hash, blob []byte) {
rawdb.WriteTrieNode(task.genBatch, common.Hash{}, path, hash, blob, s.scheme)
})
if s.scheme == rawdb.PathScheme {
// Configure the dangling node cleaner and also filter out boundary nodes
// only in the context of the path scheme. Deletion is forbidden in the
// hash scheme, as it can disrupt state completeness.
options = options.WithCleaner(func(path []byte) {
s.cleanPath(task.genBatch, common.Hash{}, path)
})
// Skip the left boundary if it's not the first range.
// Skip the right boundary if it's not the last range.
options = options.WithSkipBoundary(task.Next != (common.Hash{}), task.Last != common.MaxHash, boundaryAccountNodesGauge)
}
task.genTrie = trie.NewStackTrie(options)
for accountHash, subtasks := range task.SubTasks {
for _, subtask := range subtasks {
@@ -758,6 +782,17 @@ func (s *Syncer) loadSyncStatus() {
options = options.WithWriter(func(path []byte, hash common.Hash, blob []byte) {
rawdb.WriteTrieNode(subtask.genBatch, owner, path, hash, blob, s.scheme)
})
if s.scheme == rawdb.PathScheme {
// Configure the dangling node cleaner and also filter out boundary nodes
// only in the context of the path scheme. Deletion is forbidden in the
// hash scheme, as it can disrupt state completeness.
options = options.WithCleaner(func(path []byte) {
s.cleanPath(subtask.genBatch, owner, path)
})
// Skip the left boundary if it's not the first range.
// Skip the right boundary if it's not the last range.
options = options.WithSkipBoundary(subtask.Next != common.Hash{}, subtask.Last != common.MaxHash, boundaryStorageNodesGauge)
}
subtask.genTrie = trie.NewStackTrie(options)
}
}
@@ -814,6 +849,17 @@ func (s *Syncer) loadSyncStatus() {
options = options.WithWriter(func(path []byte, hash common.Hash, blob []byte) {
rawdb.WriteTrieNode(batch, common.Hash{}, path, hash, blob, s.scheme)
})
if s.scheme == rawdb.PathScheme {
// Configure the dangling node cleaner and also filter out boundary nodes
// only in the context of the path scheme. Deletion is forbidden in the
// hash scheme, as it can disrupt state completeness.
options = options.WithCleaner(func(path []byte) {
s.cleanPath(batch, common.Hash{}, path)
})
// Skip the left boundary if it's not the first range.
// Skip the right boundary if it's not the last range.
options = options.WithSkipBoundary(next != common.Hash{}, last != common.MaxHash, boundaryAccountNodesGauge)
}
s.tasks = append(s.tasks, &accountTask{
Next: next,
Last: last,
@@ -1968,6 +2014,7 @@ func (s *Syncer) processStorageResponse(res *storageResponse) {
if res.subTask == nil && res.mainTask.needState[j] && (i < len(res.hashes)-1 || !res.cont) {
res.mainTask.needState[j] = false
res.mainTask.pend--
smallStorageGauge.Inc(1)
}
// If the last contract was chunked, mark it as needing healing
// to avoid writing it out to disk prematurely.
@@ -2003,7 +2050,11 @@ func (s *Syncer) processStorageResponse(res *storageResponse) {
log.Debug("Chunked large contract", "initiators", len(keys), "tail", lastKey, "chunks", chunks)
}
r := newHashRange(lastKey, chunks)
if chunks == 1 {
smallStorageGauge.Inc(1)
} else {
largeStorageGauge.Inc(1)
}
// Our first task is the one that was just filled by this response.
batch := ethdb.HookedBatch{
Batch: s.db.NewBatch(),
@@ -2016,6 +2067,14 @@ func (s *Syncer) processStorageResponse(res *storageResponse) {
options = options.WithWriter(func(path []byte, hash common.Hash, blob []byte) {
rawdb.WriteTrieNode(batch, owner, path, hash, blob, s.scheme)
})
if s.scheme == rawdb.PathScheme {
options = options.WithCleaner(func(path []byte) {
s.cleanPath(batch, owner, path)
})
// Keep the left boundary as it's the first range.
// Skip the right boundary if it's not the last range.
options = options.WithSkipBoundary(false, r.End() != common.MaxHash, boundaryStorageNodesGauge)
}
tasks = append(tasks, &storageTask{
Next: common.Hash{},
Last: r.End(),
@@ -2034,6 +2093,17 @@ func (s *Syncer) processStorageResponse(res *storageResponse) {
options = options.WithWriter(func(path []byte, hash common.Hash, blob []byte) {
rawdb.WriteTrieNode(batch, owner, path, hash, blob, s.scheme)
})
if s.scheme == rawdb.PathScheme {
// Configure the dangling node cleaner and also filter out boundary nodes
// only in the context of the path scheme. Deletion is forbidden in the
// hash scheme, as it can disrupt state completeness.
options = options.WithCleaner(func(path []byte) {
s.cleanPath(batch, owner, path)
})
// Skip the left boundary as it's not the first range
// Skip the right boundary if it's not the last range.
options = options.WithSkipBoundary(true, r.End() != common.MaxHash, boundaryStorageNodesGauge)
}
tasks = append(tasks, &storageTask{
Next: r.Start(),
Last: r.End(),
@@ -2089,6 +2159,17 @@ func (s *Syncer) processStorageResponse(res *storageResponse) {
options = options.WithWriter(func(path []byte, hash common.Hash, blob []byte) {
rawdb.WriteTrieNode(batch, account, path, hash, blob, s.scheme)
})
if s.scheme == rawdb.PathScheme {
// Configure the dangling node cleaner only in the context of the
// path scheme. Deletion is forbidden in the hash scheme, as it can
// disrupt state completeness.
//
// Notably, boundary nodes can be also kept because the whole storage
// trie is complete.
options = options.WithCleaner(func(path []byte) {
s.cleanPath(batch, account, path)
})
}
tr := trie.NewStackTrie(options)
for j := 0; j < len(res.hashes[i]); j++ {
tr.Update(res.hashes[i][j][:], res.slots[i][j])
@@ -2112,16 +2193,24 @@ func (s *Syncer) processStorageResponse(res *storageResponse) {
if res.subTask != nil {
if res.subTask.done {
root := res.subTask.genTrie.Commit()
if root == res.subTask.root {
// If the chunk's root is an overflown but full delivery, clear the heal request
if err := res.subTask.genBatch.Write(); err != nil {
log.Error("Failed to persist stack slots", "err", err)
}
res.subTask.genBatch.Reset()
// If the chunk's root is an overflown but full delivery,
// clear the heal request.
accountHash := res.accounts[len(res.accounts)-1]
if root == res.subTask.root && rawdb.HasStorageTrieNode(s.db, accountHash, nil, root) {
for i, account := range res.mainTask.res.hashes {
if account == res.accounts[len(res.accounts)-1] {
if account == accountHash {
res.mainTask.needHeal[i] = false
skipStorageHealingGauge.Inc(1)
}
}
}
}
if res.subTask.genBatch.ValueSize() > ethdb.IdealBatchSize || res.subTask.done {
if res.subTask.genBatch.ValueSize() > ethdb.IdealBatchSize {
if err := res.subTask.genBatch.Write(); err != nil {
log.Error("Failed to persist stack slots", "err", err)
}