From ea2e66a58e48ef63566d5274c4a875e817a1cd39 Mon Sep 17 00:00:00 2001 From: rjl493456442 Date: Tue, 31 Oct 2023 17:39:55 +0800 Subject: [PATCH] trie/triedb/pathdb: improve dirty node flushing trigger (#28426) * trie/triedb/pathdb: improve dirty node flushing trigger * trie/triedb/pathdb: add tests * trie/triedb/pathdb: address comment --- trie/triedb/pathdb/database_test.go | 51 ++++++++++++++++++++----- trie/triedb/pathdb/disklayer.go | 58 +++++++++++++++++++++-------- trie/triedb/pathdb/history.go | 26 ++++--------- 3 files changed, 93 insertions(+), 42 deletions(-) diff --git a/trie/triedb/pathdb/database_test.go b/trie/triedb/pathdb/database_test.go index 912364f7f4..5509682c39 100644 --- a/trie/triedb/pathdb/database_test.go +++ b/trie/triedb/pathdb/database_test.go @@ -96,11 +96,15 @@ type tester struct { snapStorages map[common.Hash]map[common.Hash]map[common.Hash][]byte } -func newTester(t *testing.T) *tester { +func newTester(t *testing.T, historyLimit uint64) *tester { var ( disk, _ = rawdb.NewDatabaseWithFreezer(rawdb.NewMemoryDatabase(), t.TempDir(), "", false) - db = New(disk, &Config{CleanCacheSize: 256 * 1024, DirtyCacheSize: 256 * 1024}) - obj = &tester{ + db = New(disk, &Config{ + StateHistory: historyLimit, + CleanCacheSize: 256 * 1024, + DirtyCacheSize: 256 * 1024, + }) + obj = &tester{ db: db, preimages: make(map[common.Hash]common.Address), accounts: make(map[common.Hash][]byte), @@ -376,7 +380,7 @@ func (t *tester) bottomIndex() int { func TestDatabaseRollback(t *testing.T) { // Verify state histories - tester := newTester(t) + tester := newTester(t, 0) defer tester.release() if err := tester.verifyHistory(); err != nil { @@ -402,7 +406,7 @@ func TestDatabaseRollback(t *testing.T) { func TestDatabaseRecoverable(t *testing.T) { var ( - tester = newTester(t) + tester = newTester(t, 0) index = tester.bottomIndex() ) defer tester.release() @@ -440,7 +444,7 @@ func TestDatabaseRecoverable(t *testing.T) { } func TestDisable(t *testing.T) { - tester := newTester(t) + tester := newTester(t, 0) defer tester.release() _, stored := rawdb.ReadAccountTrieNode(tester.db.diskdb, nil) @@ -476,7 +480,7 @@ func TestDisable(t *testing.T) { } func TestCommit(t *testing.T) { - tester := newTester(t) + tester := newTester(t, 0) defer tester.release() if err := tester.db.Commit(tester.lastHash(), false); err != nil { @@ -500,7 +504,7 @@ func TestCommit(t *testing.T) { } func TestJournal(t *testing.T) { - tester := newTester(t) + tester := newTester(t, 0) defer tester.release() if err := tester.db.Journal(tester.lastHash()); err != nil { @@ -524,7 +528,7 @@ func TestJournal(t *testing.T) { } func TestCorruptedJournal(t *testing.T) { - tester := newTester(t) + tester := newTester(t, 0) defer tester.release() if err := tester.db.Journal(tester.lastHash()); err != nil { @@ -553,6 +557,35 @@ func TestCorruptedJournal(t *testing.T) { } } +// TestTailTruncateHistory function is designed to test a specific edge case where, +// when history objects are removed from the end, it should trigger a state flush +// if the ID of the new tail object is even higher than the persisted state ID. +// +// For example, let's say the ID of the persistent state is 10, and the current +// history objects range from ID(5) to ID(15). As we accumulate six more objects, +// the history will expand to cover ID(11) to ID(21). ID(11) then becomes the +// oldest history object, and its ID is even higher than the stored state. +// +// In this scenario, it is mandatory to update the persistent state before +// truncating the tail histories. This ensures that the ID of the persistent state +// always falls within the range of [oldest-history-id, latest-history-id]. +func TestTailTruncateHistory(t *testing.T) { + tester := newTester(t, 10) + defer tester.release() + + tester.db.Close() + tester.db = New(tester.db.diskdb, &Config{StateHistory: 10}) + + head, err := tester.db.freezer.Ancients() + if err != nil { + t.Fatalf("Failed to obtain freezer head") + } + stored := rawdb.ReadPersistentStateID(tester.db.diskdb) + if head != stored { + t.Fatalf("Failed to truncate excess history object above, stored: %d, head: %d", stored, head) + } +} + // copyAccounts returns a deep-copied account set of the provided one. func copyAccounts(set map[common.Hash][]byte) map[common.Hash][]byte { copied := make(map[common.Hash][]byte, len(set)) diff --git a/trie/triedb/pathdb/disklayer.go b/trie/triedb/pathdb/disklayer.go index d3b6419cc5..ef697cbce8 100644 --- a/trie/triedb/pathdb/disklayer.go +++ b/trie/triedb/pathdb/disklayer.go @@ -172,37 +172,65 @@ func (dl *diskLayer) commit(bottom *diffLayer, force bool) (*diskLayer, error) { dl.lock.Lock() defer dl.lock.Unlock() - // Construct and store the state history first. If crash happens - // after storing the state history but without flushing the - // corresponding states(journal), the stored state history will - // be truncated in the next restart. + // Construct and store the state history first. If crash happens after storing + // the state history but without flushing the corresponding states(journal), + // the stored state history will be truncated from head in the next restart. + var ( + overflow bool + oldest uint64 + ) if dl.db.freezer != nil { - err := writeHistory(dl.db.diskdb, dl.db.freezer, bottom, dl.db.config.StateHistory) + err := writeHistory(dl.db.freezer, bottom) if err != nil { return nil, err } + // Determine if the persisted history object has exceeded the configured + // limitation, set the overflow as true if so. + tail, err := dl.db.freezer.Tail() + if err != nil { + return nil, err + } + limit := dl.db.config.StateHistory + if limit != 0 && bottom.stateID()-tail > limit { + overflow = true + oldest = bottom.stateID() - limit + 1 // track the id of history **after truncation** + } } // Mark the diskLayer as stale before applying any mutations on top. dl.stale = true - // Store the root->id lookup afterwards. All stored lookups are - // identified by the **unique** state root. It's impossible that - // in the same chain blocks are not adjacent but have the same - // root. + // Store the root->id lookup afterwards. All stored lookups are identified + // by the **unique** state root. It's impossible that in the same chain + // blocks are not adjacent but have the same root. if dl.id == 0 { rawdb.WriteStateID(dl.db.diskdb, dl.root, 0) } rawdb.WriteStateID(dl.db.diskdb, bottom.rootHash(), bottom.stateID()) - // Construct a new disk layer by merging the nodes from the provided - // diff layer, and flush the content in disk layer if there are too - // many nodes cached. The clean cache is inherited from the original - // disk layer for reusing. + // Construct a new disk layer by merging the nodes from the provided diff + // layer, and flush the content in disk layer if there are too many nodes + // cached. The clean cache is inherited from the original disk layer. ndl := newDiskLayer(bottom.root, bottom.stateID(), dl.db, dl.cleans, dl.buffer.commit(bottom.nodes)) - err := ndl.buffer.flush(ndl.db.diskdb, ndl.cleans, ndl.id, force) - if err != nil { + + // In a unique scenario where the ID of the oldest history object (after tail + // truncation) surpasses the persisted state ID, we take the necessary action + // of forcibly committing the cached dirty nodes to ensure that the persisted + // state ID remains higher. + if !force && rawdb.ReadPersistentStateID(dl.db.diskdb) < oldest { + force = true + } + if err := ndl.buffer.flush(ndl.db.diskdb, ndl.cleans, ndl.id, force); err != nil { return nil, err } + // To remove outdated history objects from the end, we set the 'tail' parameter + // to 'oldest-1' due to the offset between the freezer index and the history ID. + if overflow { + pruned, err := truncateFromTail(ndl.db.diskdb, ndl.db.freezer, oldest-1) + if err != nil { + return nil, err + } + log.Debug("Pruned state history", "items", pruned, "tailid", oldest) + } return ndl, nil } diff --git a/trie/triedb/pathdb/history.go b/trie/triedb/pathdb/history.go index 6f33b61859..6e3f3faaed 100644 --- a/trie/triedb/pathdb/history.go +++ b/trie/triedb/pathdb/history.go @@ -512,38 +512,28 @@ func readHistory(freezer *rawdb.ResettableFreezer, id uint64) (*history, error) return &dec, nil } -// writeHistory writes the state history with provided state set. After -// storing the corresponding state history, it will also prune the stale -// histories from the disk with the given threshold. -func writeHistory(db ethdb.KeyValueStore, freezer *rawdb.ResettableFreezer, dl *diffLayer, limit uint64) error { +// writeHistory persists the state history with the provided state set. +func writeHistory(freezer *rawdb.ResettableFreezer, dl *diffLayer) error { // Short circuit if state set is not available. if dl.states == nil { return errors.New("state change set is not available") } var ( - err error - n int - start = time.Now() - h = newHistory(dl.rootHash(), dl.parentLayer().rootHash(), dl.block, dl.states) + start = time.Now() + history = newHistory(dl.rootHash(), dl.parentLayer().rootHash(), dl.block, dl.states) ) - accountData, storageData, accountIndex, storageIndex := h.encode() + accountData, storageData, accountIndex, storageIndex := history.encode() dataSize := common.StorageSize(len(accountData) + len(storageData)) indexSize := common.StorageSize(len(accountIndex) + len(storageIndex)) // Write history data into five freezer table respectively. - rawdb.WriteStateHistory(freezer, dl.stateID(), h.meta.encode(), accountIndex, storageIndex, accountData, storageData) + rawdb.WriteStateHistory(freezer, dl.stateID(), history.meta.encode(), accountIndex, storageIndex, accountData, storageData) - // Prune stale state histories based on the config. - if limit != 0 && dl.stateID() > limit { - n, err = truncateFromTail(db, freezer, dl.stateID()-limit) - if err != nil { - return err - } - } historyDataBytesMeter.Mark(int64(dataSize)) historyIndexBytesMeter.Mark(int64(indexSize)) historyBuildTimeMeter.UpdateSince(start) - log.Debug("Stored state history", "id", dl.stateID(), "block", dl.block, "data", dataSize, "index", indexSize, "pruned", n, "elapsed", common.PrettyDuration(time.Since(start))) + log.Debug("Stored state history", "id", dl.stateID(), "block", dl.block, "data", dataSize, "index", indexSize, "elapsed", common.PrettyDuration(time.Since(start))) + return nil }