go-ethereum/core/state/sync_test.go

488 lines
16 KiB
Go
Raw Normal View History

// Copyright 2015 The go-ethereum Authors
// This file is part of the go-ethereum library.
//
// The go-ethereum library is free software: you can redistribute it and/or modify
// it under the terms of the GNU Lesser General Public License as published by
// the Free Software Foundation, either version 3 of the License, or
// (at your option) any later version.
//
// The go-ethereum library is distributed in the hope that it will be useful,
// but WITHOUT ANY WARRANTY; without even the implied warranty of
// MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
// GNU Lesser General Public License for more details.
//
// You should have received a copy of the GNU Lesser General Public License
// along with the go-ethereum library. If not, see <http://www.gnu.org/licenses/>.
package state
import (
"bytes"
"math/big"
"testing"
"github.com/ethereum/go-ethereum/common"
"github.com/ethereum/go-ethereum/core/rawdb"
"github.com/ethereum/go-ethereum/crypto"
"github.com/ethereum/go-ethereum/ethdb"
"github.com/ethereum/go-ethereum/ethdb/memorydb"
"github.com/ethereum/go-ethereum/rlp"
"github.com/ethereum/go-ethereum/trie"
)
// testAccount is the data associated with an account used by the state tests.
type testAccount struct {
address common.Address
balance *big.Int
nonce uint64
code []byte
}
// makeTestState create a sample test state to test node-wise reconstruction.
func makeTestState() (Database, common.Hash, []*testAccount) {
// Create an empty state
db := NewDatabase(rawdb.NewMemoryDatabase())
state, _ := New(common.Hash{}, db, nil)
// Fill it with some arbitrary data
var accounts []*testAccount
for i := byte(0); i < 96; i++ {
obj := state.GetOrNewStateObject(common.BytesToAddress([]byte{i}))
acc := &testAccount{address: common.BytesToAddress([]byte{i})}
obj.AddBalance(big.NewInt(int64(11 * i)))
acc.balance = big.NewInt(int64(11 * i))
obj.SetNonce(uint64(42 * i))
acc.nonce = uint64(42 * i)
if i%3 == 0 {
obj.SetCode(crypto.Keccak256Hash([]byte{i, i, i, i, i}), []byte{i, i, i, i, i})
acc.code = []byte{i, i, i, i, i}
}
if i%5 == 0 {
for j := byte(0); j < 5; j++ {
hash := crypto.Keccak256Hash([]byte{i, i, i, i, i, j, j})
obj.SetState(db, hash, hash)
}
}
state.updateStateObject(obj)
accounts = append(accounts, acc)
}
root, _ := state.Commit(false)
// Return the generated state
return db, root, accounts
}
// checkStateAccounts cross references a reconstructed state with an expected
// account array.
func checkStateAccounts(t *testing.T, db ethdb.Database, root common.Hash, accounts []*testAccount) {
// Check root availability and state contents
state, err := New(root, NewDatabase(db), nil)
if err != nil {
t.Fatalf("failed to create state trie at %x: %v", root, err)
}
if err := checkStateConsistency(db, root); err != nil {
t.Fatalf("inconsistent state trie at %x: %v", root, err)
}
for i, acc := range accounts {
if balance := state.GetBalance(acc.address); balance.Cmp(acc.balance) != 0 {
t.Errorf("account %d: balance mismatch: have %v, want %v", i, balance, acc.balance)
}
if nonce := state.GetNonce(acc.address); nonce != acc.nonce {
t.Errorf("account %d: nonce mismatch: have %v, want %v", i, nonce, acc.nonce)
}
if code := state.GetCode(acc.address); !bytes.Equal(code, acc.code) {
t.Errorf("account %d: code mismatch: have %x, want %x", i, code, acc.code)
}
}
}
// checkTrieConsistency checks that all nodes in a (sub-)trie are indeed present.
func checkTrieConsistency(db ethdb.Database, root common.Hash) error {
if v, _ := db.Get(root[:]); v == nil {
return nil // Consider a non existent state consistent.
}
trie, err := trie.New(root, trie.NewDatabase(db))
if err != nil {
return err
}
it := trie.NodeIterator(nil)
for it.Next(true) {
}
return it.Error()
}
// checkStateConsistency checks that all data of a state root is present.
func checkStateConsistency(db ethdb.Database, root common.Hash) error {
// Create and iterate a state trie rooted in a sub-node
if _, err := db.Get(root.Bytes()); err != nil {
return nil // Consider a non existent state consistent.
}
state, err := New(root, NewDatabase(db), nil)
if err != nil {
return err
}
it := NewNodeIterator(state)
for it.Next() {
}
return it.Error
}
// Tests that an empty state is not scheduled for syncing.
func TestEmptyStateSync(t *testing.T) {
empty := common.HexToHash("56e81f171bcc55a6ff8345e692c0f86e5b48e01b996cadc001622fb5e363b421")
core, eth: faster snapshot generation (#22504) * eth/protocols: persist received state segments * core: initial implementation * core/state/snapshot: add tests * core, eth: updates * eth/protocols/snapshot: count flat state size * core/state: add metrics * core/state/snapshot: skip unnecessary deletion * core/state/snapshot: rename * core/state/snapshot: use the global batch * core/state/snapshot: add logs and fix wiping * core/state/snapshot: fix * core/state/snapshot: save generation progress even if the batch is empty * core/state/snapshot: fixes * core/state/snapshot: fix initial account range length * core/state/snapshot: fix initial account range * eth/protocols/snap: store flat states during the healing * eth/protocols/snap: print logs * core/state/snapshot: refactor (#4) * core/state/snapshot: refactor * core/state/snapshot: tiny fix and polish Co-authored-by: rjl493456442 <garyrong0905@gmail.com> * core, eth: fixes * core, eth: fix healing writer * core, trie, eth: fix paths * eth/protocols/snap: fix encoding * eth, core: add debug log * core/state/generate: release iterator asap (#5) core/state/snapshot: less copy core/state/snapshot: revert split loop core/state/snapshot: handle storage becoming empty, improve test robustness core/state: test modified codehash core/state/snapshot: polish * core/state/snapshot: optimize stats counter * core, eth: add metric * core/state/snapshot: update comments * core/state/snapshot: improve tests * core/state/snapshot: replace secure trie with standard trie * core/state/snapshot: wrap return as the struct * core/state/snapshot: skip wiping correct states * core/state/snapshot: updates * core/state/snapshot: fixes * core/state/snapshot: fix panic due to reference flaw in closure * core/state/snapshot: fix errors in state generation logic + fix log output * core/state/snapshot: remove an error case * core/state/snapshot: fix condition-check for exhausted snap state * core/state/snapshot: use stackTrie for small tries * core/state/snapshot: don't resolve small storage tries in vain * core/state/snapshot: properly clean up storage of deleted accounts * core/state/snapshot: avoid RLP-encoding in some cases + minor nitpicks * core/state/snapshot: fix error (+testcase) * core/state/snapshot: clean up tests a bit * core/state/snapshot: work in progress on better tests * core/state/snapshot: polish code * core/state/snapshot: fix trie iteration abortion trigger * core/state/snapshot: fixes flaws * core/state/snapshot: remove panic * core/state/snapshot: fix abort * core/state/snapshot: more tests (plus failing testcase) * core/state/snapshot: more testcases + fix for failing test * core/state/snapshot: testcase for malformed data * core/state/snapshot: some test nitpicks * core/state/snapshot: improvements to logging * core/state/snapshot: testcase to demo error in abortion * core/state/snapshot: fix abortion * cmd/geth: make verify-state report the root * trie: fix failing test * core/state/snapshot: add timer metrics * core/state/snapshot: fix metrics * core/state/snapshot: udpate tests * eth/protocols/snap: write snapshot account even if code or state is needed * core/state/snapshot: fix diskmore check * core/state/snapshot: review fixes * core/state/snapshot: improve error message * cmd/geth: rename 'error' to 'err' in logs * core/state/snapshot: fix some review concerns * core/state/snapshot, eth/protocols/snap: clear snapshot marker when starting/resuming snap sync * core: add error log * core/state/snapshot: use proper timers for metrics collection * core/state/snapshot: address some review concerns * eth/protocols/snap: improved log message * eth/protocols/snap: fix heal logs to condense infos * core/state/snapshot: wait for generator termination before restarting * core/state/snapshot: revert timers to counters to track total time Co-authored-by: Martin Holst Swende <martin@swende.se> Co-authored-by: Péter Szilágyi <peterke@gmail.com>
2021-04-14 23:23:11 +03:00
sync := NewStateSync(empty, rawdb.NewMemoryDatabase(), trie.NewSyncBloom(1, memorydb.New()), nil)
if nodes, paths, codes := sync.Missing(1); len(nodes) != 0 || len(paths) != 0 || len(codes) != 0 {
t.Errorf(" content requested for empty state: %v, %v, %v", nodes, paths, codes)
}
}
// Tests that given a root hash, a state can sync iteratively on a single thread,
// requesting retrieval tasks and returning all of them in one go.
func TestIterativeStateSyncIndividual(t *testing.T) {
testIterativeStateSync(t, 1, false, false)
}
func TestIterativeStateSyncBatched(t *testing.T) {
testIterativeStateSync(t, 100, false, false)
}
func TestIterativeStateSyncIndividualFromDisk(t *testing.T) {
testIterativeStateSync(t, 1, true, false)
}
func TestIterativeStateSyncBatchedFromDisk(t *testing.T) {
testIterativeStateSync(t, 100, true, false)
}
func TestIterativeStateSyncIndividualByPath(t *testing.T) {
testIterativeStateSync(t, 1, false, true)
}
func TestIterativeStateSyncBatchedByPath(t *testing.T) {
testIterativeStateSync(t, 100, false, true)
}
func testIterativeStateSync(t *testing.T, count int, commit bool, bypath bool) {
// Create a random state to copy
srcDb, srcRoot, srcAccounts := makeTestState()
if commit {
srcDb.TrieDB().Commit(srcRoot, false, nil)
}
srcTrie, _ := trie.New(srcRoot, srcDb.TrieDB())
// Create a destination state and sync with the scheduler
dstDb := rawdb.NewMemoryDatabase()
core, eth: faster snapshot generation (#22504) * eth/protocols: persist received state segments * core: initial implementation * core/state/snapshot: add tests * core, eth: updates * eth/protocols/snapshot: count flat state size * core/state: add metrics * core/state/snapshot: skip unnecessary deletion * core/state/snapshot: rename * core/state/snapshot: use the global batch * core/state/snapshot: add logs and fix wiping * core/state/snapshot: fix * core/state/snapshot: save generation progress even if the batch is empty * core/state/snapshot: fixes * core/state/snapshot: fix initial account range length * core/state/snapshot: fix initial account range * eth/protocols/snap: store flat states during the healing * eth/protocols/snap: print logs * core/state/snapshot: refactor (#4) * core/state/snapshot: refactor * core/state/snapshot: tiny fix and polish Co-authored-by: rjl493456442 <garyrong0905@gmail.com> * core, eth: fixes * core, eth: fix healing writer * core, trie, eth: fix paths * eth/protocols/snap: fix encoding * eth, core: add debug log * core/state/generate: release iterator asap (#5) core/state/snapshot: less copy core/state/snapshot: revert split loop core/state/snapshot: handle storage becoming empty, improve test robustness core/state: test modified codehash core/state/snapshot: polish * core/state/snapshot: optimize stats counter * core, eth: add metric * core/state/snapshot: update comments * core/state/snapshot: improve tests * core/state/snapshot: replace secure trie with standard trie * core/state/snapshot: wrap return as the struct * core/state/snapshot: skip wiping correct states * core/state/snapshot: updates * core/state/snapshot: fixes * core/state/snapshot: fix panic due to reference flaw in closure * core/state/snapshot: fix errors in state generation logic + fix log output * core/state/snapshot: remove an error case * core/state/snapshot: fix condition-check for exhausted snap state * core/state/snapshot: use stackTrie for small tries * core/state/snapshot: don't resolve small storage tries in vain * core/state/snapshot: properly clean up storage of deleted accounts * core/state/snapshot: avoid RLP-encoding in some cases + minor nitpicks * core/state/snapshot: fix error (+testcase) * core/state/snapshot: clean up tests a bit * core/state/snapshot: work in progress on better tests * core/state/snapshot: polish code * core/state/snapshot: fix trie iteration abortion trigger * core/state/snapshot: fixes flaws * core/state/snapshot: remove panic * core/state/snapshot: fix abort * core/state/snapshot: more tests (plus failing testcase) * core/state/snapshot: more testcases + fix for failing test * core/state/snapshot: testcase for malformed data * core/state/snapshot: some test nitpicks * core/state/snapshot: improvements to logging * core/state/snapshot: testcase to demo error in abortion * core/state/snapshot: fix abortion * cmd/geth: make verify-state report the root * trie: fix failing test * core/state/snapshot: add timer metrics * core/state/snapshot: fix metrics * core/state/snapshot: udpate tests * eth/protocols/snap: write snapshot account even if code or state is needed * core/state/snapshot: fix diskmore check * core/state/snapshot: review fixes * core/state/snapshot: improve error message * cmd/geth: rename 'error' to 'err' in logs * core/state/snapshot: fix some review concerns * core/state/snapshot, eth/protocols/snap: clear snapshot marker when starting/resuming snap sync * core: add error log * core/state/snapshot: use proper timers for metrics collection * core/state/snapshot: address some review concerns * eth/protocols/snap: improved log message * eth/protocols/snap: fix heal logs to condense infos * core/state/snapshot: wait for generator termination before restarting * core/state/snapshot: revert timers to counters to track total time Co-authored-by: Martin Holst Swende <martin@swende.se> Co-authored-by: Péter Szilágyi <peterke@gmail.com>
2021-04-14 23:23:11 +03:00
sched := NewStateSync(srcRoot, dstDb, trie.NewSyncBloom(1, dstDb), nil)
nodes, paths, codes := sched.Missing(count)
var (
hashQueue []common.Hash
pathQueue []trie.SyncPath
)
if !bypath {
hashQueue = append(append(hashQueue[:0], nodes...), codes...)
} else {
hashQueue = append(hashQueue[:0], codes...)
pathQueue = append(pathQueue[:0], paths...)
}
for len(hashQueue)+len(pathQueue) > 0 {
results := make([]trie.SyncResult, len(hashQueue)+len(pathQueue))
for i, hash := range hashQueue {
data, err := srcDb.TrieDB().Node(hash)
if err != nil {
data, err = srcDb.ContractCode(common.Hash{}, hash)
}
if err != nil {
t.Fatalf("failed to retrieve node data for hash %x", hash)
}
2016-04-15 12:06:57 +03:00
results[i] = trie.SyncResult{Hash: hash, Data: data}
}
for i, path := range pathQueue {
if len(path) == 1 {
data, _, err := srcTrie.TryGetNode(path[0])
if err != nil {
t.Fatalf("failed to retrieve node data for path %x: %v", path, err)
}
results[len(hashQueue)+i] = trie.SyncResult{Hash: crypto.Keccak256Hash(data), Data: data}
} else {
var acc Account
if err := rlp.DecodeBytes(srcTrie.Get(path[0]), &acc); err != nil {
t.Fatalf("failed to decode account on path %x: %v", path, err)
}
stTrie, err := trie.New(acc.Root, srcDb.TrieDB())
if err != nil {
t.Fatalf("failed to retriev storage trie for path %x: %v", path, err)
}
data, _, err := stTrie.TryGetNode(path[1])
if err != nil {
t.Fatalf("failed to retrieve node data for path %x: %v", path, err)
}
results[len(hashQueue)+i] = trie.SyncResult{Hash: crypto.Keccak256Hash(data), Data: data}
}
}
for _, result := range results {
if err := sched.Process(result); err != nil {
t.Errorf("failed to process result %v", err)
}
}
batch := dstDb.NewBatch()
if err := sched.Commit(batch); err != nil {
t.Fatalf("failed to commit data: %v", err)
eth/downloader: separate state sync from queue (#14460) * eth/downloader: separate state sync from queue Scheduling of state node downloads hogged the downloader queue lock when new requests were scheduled. This caused timeouts for other requests. With this change, state sync is fully independent of all other downloads and doesn't involve the queue at all. State sync is started and checked on in processContent. This is slightly awkward because processContent doesn't have a select loop. Instead, the queue is closed by an auxiliary goroutine when state sync fails. We tried several alternatives to this but settled on the current approach because it's the least amount of change overall. Handling of the pivot block has changed slightly: the queue previously prevented import of pivot block receipts before the state of the pivot block was available. In this commit, the receipt will be imported before the state. This causes an annoyance where the pivot block is committed as fast block head even when state downloads fail. Stay tuned for more updates in this area ;) * eth/downloader: remove cancelTimeout channel * eth/downloader: retry state requests on timeout * eth/downloader: improve comment * eth/downloader: mark peers idle when state sync is done * eth/downloader: move pivot block splitting to processContent This change also ensures that pivot block receipts aren't imported before the pivot block itself. * eth/downloader: limit state node retries * eth/downloader: improve state node error handling and retry check * eth/downloader: remove maxStateNodeRetries It fails the sync too much. * eth/downloader: remove last use of cancelCh in statesync.go Fixes TestDeliverHeadersHang*Fast and (hopefully) the weird cancellation behaviour at the end of fast sync. * eth/downloader: fix leak in runStateSync * eth/downloader: don't run processFullSyncContent in LightSync mode * eth/downloader: improve comments * eth/downloader: fix vet, megacheck * eth/downloader: remove unrequested tasks anyway * eth/downloader, trie: various polishes around duplicate items This commit explicitly tracks duplicate and unexpected state delieveries done against a trie Sync structure, also adding there to import info logs. The commit moves the db batch used to commit trie changes one level deeper so its flushed after every node insertion. This is needed to avoid a lot of duplicate retrievals caused by inconsistencies between Sync internals and database. A better approach is to track not-yet-written states in trie.Sync and flush on commit, but I'm focuing on correctness first now. The commit fixes a regression around pivot block fail count. The counter previously was reset to 1 if and only if a sync cycle progressed (inserted at least 1 entry to the database). The current code reset it already if a node was delivered, which is not stong enough, because unless it ends up written to disk, an attacker can just loop and attack ad infinitum. The commit also fixes a regression around state deliveries and timeouts. The old downloader tracked if a delivery is stale (none of the deliveries were requestedt), in which case it didn't mark the node idle and did not send further requests, since it signals a past timeout. The current code did mark it idle even on stale deliveries, which eventually caused two requests to be in flight at the same time, making the deliveries always stale and mass duplicating retrievals between multiple peers. * eth/downloader: fix state request leak This commit fixes the hang seen sometimes while doing the state sync. The cause of the hang was a rare combination of events: request state data from peer, peer drops and reconnects almost immediately. This caused a new download task to be assigned to the peer, overwriting the old one still waiting for a timeout, which in turned leaked the requests out, never to be retried. The fix is to ensure that a task assignment moves any pending one back into the retry queue. The commit also fixes a regression with peer dropping due to stalls. The current code considered a peer stalling if they timed out delivering 1 item. However, the downloader never requests only one, the minimum is 2 (attempt to fine tune estimated latency/bandwidth). The fix is simply to drop if a timeout is detected at 2 items. Apart from the above bugfixes, the commit contains some code polishes I made while debugging the hang. * core, eth, trie: support batched trie sync db writes * trie: rename SyncMemCache to syncMemBatch
2017-06-22 15:26:03 +03:00
}
batch.Write()
nodes, paths, codes = sched.Missing(count)
if !bypath {
hashQueue = append(append(hashQueue[:0], nodes...), codes...)
} else {
hashQueue = append(hashQueue[:0], codes...)
pathQueue = append(pathQueue[:0], paths...)
}
}
// Cross check that the two states are in sync
checkStateAccounts(t, dstDb, srcRoot, srcAccounts)
}
// Tests that the trie scheduler can correctly reconstruct the state even if only
// partial results are returned, and the others sent only later.
func TestIterativeDelayedStateSync(t *testing.T) {
// Create a random state to copy
srcDb, srcRoot, srcAccounts := makeTestState()
// Create a destination state and sync with the scheduler
dstDb := rawdb.NewMemoryDatabase()
core, eth: faster snapshot generation (#22504) * eth/protocols: persist received state segments * core: initial implementation * core/state/snapshot: add tests * core, eth: updates * eth/protocols/snapshot: count flat state size * core/state: add metrics * core/state/snapshot: skip unnecessary deletion * core/state/snapshot: rename * core/state/snapshot: use the global batch * core/state/snapshot: add logs and fix wiping * core/state/snapshot: fix * core/state/snapshot: save generation progress even if the batch is empty * core/state/snapshot: fixes * core/state/snapshot: fix initial account range length * core/state/snapshot: fix initial account range * eth/protocols/snap: store flat states during the healing * eth/protocols/snap: print logs * core/state/snapshot: refactor (#4) * core/state/snapshot: refactor * core/state/snapshot: tiny fix and polish Co-authored-by: rjl493456442 <garyrong0905@gmail.com> * core, eth: fixes * core, eth: fix healing writer * core, trie, eth: fix paths * eth/protocols/snap: fix encoding * eth, core: add debug log * core/state/generate: release iterator asap (#5) core/state/snapshot: less copy core/state/snapshot: revert split loop core/state/snapshot: handle storage becoming empty, improve test robustness core/state: test modified codehash core/state/snapshot: polish * core/state/snapshot: optimize stats counter * core, eth: add metric * core/state/snapshot: update comments * core/state/snapshot: improve tests * core/state/snapshot: replace secure trie with standard trie * core/state/snapshot: wrap return as the struct * core/state/snapshot: skip wiping correct states * core/state/snapshot: updates * core/state/snapshot: fixes * core/state/snapshot: fix panic due to reference flaw in closure * core/state/snapshot: fix errors in state generation logic + fix log output * core/state/snapshot: remove an error case * core/state/snapshot: fix condition-check for exhausted snap state * core/state/snapshot: use stackTrie for small tries * core/state/snapshot: don't resolve small storage tries in vain * core/state/snapshot: properly clean up storage of deleted accounts * core/state/snapshot: avoid RLP-encoding in some cases + minor nitpicks * core/state/snapshot: fix error (+testcase) * core/state/snapshot: clean up tests a bit * core/state/snapshot: work in progress on better tests * core/state/snapshot: polish code * core/state/snapshot: fix trie iteration abortion trigger * core/state/snapshot: fixes flaws * core/state/snapshot: remove panic * core/state/snapshot: fix abort * core/state/snapshot: more tests (plus failing testcase) * core/state/snapshot: more testcases + fix for failing test * core/state/snapshot: testcase for malformed data * core/state/snapshot: some test nitpicks * core/state/snapshot: improvements to logging * core/state/snapshot: testcase to demo error in abortion * core/state/snapshot: fix abortion * cmd/geth: make verify-state report the root * trie: fix failing test * core/state/snapshot: add timer metrics * core/state/snapshot: fix metrics * core/state/snapshot: udpate tests * eth/protocols/snap: write snapshot account even if code or state is needed * core/state/snapshot: fix diskmore check * core/state/snapshot: review fixes * core/state/snapshot: improve error message * cmd/geth: rename 'error' to 'err' in logs * core/state/snapshot: fix some review concerns * core/state/snapshot, eth/protocols/snap: clear snapshot marker when starting/resuming snap sync * core: add error log * core/state/snapshot: use proper timers for metrics collection * core/state/snapshot: address some review concerns * eth/protocols/snap: improved log message * eth/protocols/snap: fix heal logs to condense infos * core/state/snapshot: wait for generator termination before restarting * core/state/snapshot: revert timers to counters to track total time Co-authored-by: Martin Holst Swende <martin@swende.se> Co-authored-by: Péter Szilágyi <peterke@gmail.com>
2021-04-14 23:23:11 +03:00
sched := NewStateSync(srcRoot, dstDb, trie.NewSyncBloom(1, dstDb), nil)
nodes, _, codes := sched.Missing(0)
queue := append(append([]common.Hash{}, nodes...), codes...)
for len(queue) > 0 {
// Sync only half of the scheduled nodes
results := make([]trie.SyncResult, len(queue)/2+1)
for i, hash := range queue[:len(results)] {
data, err := srcDb.TrieDB().Node(hash)
if err != nil {
data, err = srcDb.ContractCode(common.Hash{}, hash)
}
if err != nil {
t.Fatalf("failed to retrieve node data for %x", hash)
}
2016-04-15 12:06:57 +03:00
results[i] = trie.SyncResult{Hash: hash, Data: data}
}
for _, result := range results {
if err := sched.Process(result); err != nil {
t.Fatalf("failed to process result %v", err)
}
}
batch := dstDb.NewBatch()
if err := sched.Commit(batch); err != nil {
t.Fatalf("failed to commit data: %v", err)
eth/downloader: separate state sync from queue (#14460) * eth/downloader: separate state sync from queue Scheduling of state node downloads hogged the downloader queue lock when new requests were scheduled. This caused timeouts for other requests. With this change, state sync is fully independent of all other downloads and doesn't involve the queue at all. State sync is started and checked on in processContent. This is slightly awkward because processContent doesn't have a select loop. Instead, the queue is closed by an auxiliary goroutine when state sync fails. We tried several alternatives to this but settled on the current approach because it's the least amount of change overall. Handling of the pivot block has changed slightly: the queue previously prevented import of pivot block receipts before the state of the pivot block was available. In this commit, the receipt will be imported before the state. This causes an annoyance where the pivot block is committed as fast block head even when state downloads fail. Stay tuned for more updates in this area ;) * eth/downloader: remove cancelTimeout channel * eth/downloader: retry state requests on timeout * eth/downloader: improve comment * eth/downloader: mark peers idle when state sync is done * eth/downloader: move pivot block splitting to processContent This change also ensures that pivot block receipts aren't imported before the pivot block itself. * eth/downloader: limit state node retries * eth/downloader: improve state node error handling and retry check * eth/downloader: remove maxStateNodeRetries It fails the sync too much. * eth/downloader: remove last use of cancelCh in statesync.go Fixes TestDeliverHeadersHang*Fast and (hopefully) the weird cancellation behaviour at the end of fast sync. * eth/downloader: fix leak in runStateSync * eth/downloader: don't run processFullSyncContent in LightSync mode * eth/downloader: improve comments * eth/downloader: fix vet, megacheck * eth/downloader: remove unrequested tasks anyway * eth/downloader, trie: various polishes around duplicate items This commit explicitly tracks duplicate and unexpected state delieveries done against a trie Sync structure, also adding there to import info logs. The commit moves the db batch used to commit trie changes one level deeper so its flushed after every node insertion. This is needed to avoid a lot of duplicate retrievals caused by inconsistencies between Sync internals and database. A better approach is to track not-yet-written states in trie.Sync and flush on commit, but I'm focuing on correctness first now. The commit fixes a regression around pivot block fail count. The counter previously was reset to 1 if and only if a sync cycle progressed (inserted at least 1 entry to the database). The current code reset it already if a node was delivered, which is not stong enough, because unless it ends up written to disk, an attacker can just loop and attack ad infinitum. The commit also fixes a regression around state deliveries and timeouts. The old downloader tracked if a delivery is stale (none of the deliveries were requestedt), in which case it didn't mark the node idle and did not send further requests, since it signals a past timeout. The current code did mark it idle even on stale deliveries, which eventually caused two requests to be in flight at the same time, making the deliveries always stale and mass duplicating retrievals between multiple peers. * eth/downloader: fix state request leak This commit fixes the hang seen sometimes while doing the state sync. The cause of the hang was a rare combination of events: request state data from peer, peer drops and reconnects almost immediately. This caused a new download task to be assigned to the peer, overwriting the old one still waiting for a timeout, which in turned leaked the requests out, never to be retried. The fix is to ensure that a task assignment moves any pending one back into the retry queue. The commit also fixes a regression with peer dropping due to stalls. The current code considered a peer stalling if they timed out delivering 1 item. However, the downloader never requests only one, the minimum is 2 (attempt to fine tune estimated latency/bandwidth). The fix is simply to drop if a timeout is detected at 2 items. Apart from the above bugfixes, the commit contains some code polishes I made while debugging the hang. * core, eth, trie: support batched trie sync db writes * trie: rename SyncMemCache to syncMemBatch
2017-06-22 15:26:03 +03:00
}
batch.Write()
nodes, _, codes = sched.Missing(0)
queue = append(append(queue[len(results):], nodes...), codes...)
}
// Cross check that the two states are in sync
checkStateAccounts(t, dstDb, srcRoot, srcAccounts)
}
// Tests that given a root hash, a trie can sync iteratively on a single thread,
// requesting retrieval tasks and returning all of them in one go, however in a
// random order.
func TestIterativeRandomStateSyncIndividual(t *testing.T) { testIterativeRandomStateSync(t, 1) }
func TestIterativeRandomStateSyncBatched(t *testing.T) { testIterativeRandomStateSync(t, 100) }
func testIterativeRandomStateSync(t *testing.T, count int) {
// Create a random state to copy
srcDb, srcRoot, srcAccounts := makeTestState()
// Create a destination state and sync with the scheduler
dstDb := rawdb.NewMemoryDatabase()
core, eth: faster snapshot generation (#22504) * eth/protocols: persist received state segments * core: initial implementation * core/state/snapshot: add tests * core, eth: updates * eth/protocols/snapshot: count flat state size * core/state: add metrics * core/state/snapshot: skip unnecessary deletion * core/state/snapshot: rename * core/state/snapshot: use the global batch * core/state/snapshot: add logs and fix wiping * core/state/snapshot: fix * core/state/snapshot: save generation progress even if the batch is empty * core/state/snapshot: fixes * core/state/snapshot: fix initial account range length * core/state/snapshot: fix initial account range * eth/protocols/snap: store flat states during the healing * eth/protocols/snap: print logs * core/state/snapshot: refactor (#4) * core/state/snapshot: refactor * core/state/snapshot: tiny fix and polish Co-authored-by: rjl493456442 <garyrong0905@gmail.com> * core, eth: fixes * core, eth: fix healing writer * core, trie, eth: fix paths * eth/protocols/snap: fix encoding * eth, core: add debug log * core/state/generate: release iterator asap (#5) core/state/snapshot: less copy core/state/snapshot: revert split loop core/state/snapshot: handle storage becoming empty, improve test robustness core/state: test modified codehash core/state/snapshot: polish * core/state/snapshot: optimize stats counter * core, eth: add metric * core/state/snapshot: update comments * core/state/snapshot: improve tests * core/state/snapshot: replace secure trie with standard trie * core/state/snapshot: wrap return as the struct * core/state/snapshot: skip wiping correct states * core/state/snapshot: updates * core/state/snapshot: fixes * core/state/snapshot: fix panic due to reference flaw in closure * core/state/snapshot: fix errors in state generation logic + fix log output * core/state/snapshot: remove an error case * core/state/snapshot: fix condition-check for exhausted snap state * core/state/snapshot: use stackTrie for small tries * core/state/snapshot: don't resolve small storage tries in vain * core/state/snapshot: properly clean up storage of deleted accounts * core/state/snapshot: avoid RLP-encoding in some cases + minor nitpicks * core/state/snapshot: fix error (+testcase) * core/state/snapshot: clean up tests a bit * core/state/snapshot: work in progress on better tests * core/state/snapshot: polish code * core/state/snapshot: fix trie iteration abortion trigger * core/state/snapshot: fixes flaws * core/state/snapshot: remove panic * core/state/snapshot: fix abort * core/state/snapshot: more tests (plus failing testcase) * core/state/snapshot: more testcases + fix for failing test * core/state/snapshot: testcase for malformed data * core/state/snapshot: some test nitpicks * core/state/snapshot: improvements to logging * core/state/snapshot: testcase to demo error in abortion * core/state/snapshot: fix abortion * cmd/geth: make verify-state report the root * trie: fix failing test * core/state/snapshot: add timer metrics * core/state/snapshot: fix metrics * core/state/snapshot: udpate tests * eth/protocols/snap: write snapshot account even if code or state is needed * core/state/snapshot: fix diskmore check * core/state/snapshot: review fixes * core/state/snapshot: improve error message * cmd/geth: rename 'error' to 'err' in logs * core/state/snapshot: fix some review concerns * core/state/snapshot, eth/protocols/snap: clear snapshot marker when starting/resuming snap sync * core: add error log * core/state/snapshot: use proper timers for metrics collection * core/state/snapshot: address some review concerns * eth/protocols/snap: improved log message * eth/protocols/snap: fix heal logs to condense infos * core/state/snapshot: wait for generator termination before restarting * core/state/snapshot: revert timers to counters to track total time Co-authored-by: Martin Holst Swende <martin@swende.se> Co-authored-by: Péter Szilágyi <peterke@gmail.com>
2021-04-14 23:23:11 +03:00
sched := NewStateSync(srcRoot, dstDb, trie.NewSyncBloom(1, dstDb), nil)
queue := make(map[common.Hash]struct{})
nodes, _, codes := sched.Missing(count)
for _, hash := range append(nodes, codes...) {
queue[hash] = struct{}{}
}
for len(queue) > 0 {
// Fetch all the queued nodes in a random order
results := make([]trie.SyncResult, 0, len(queue))
2017-01-06 17:52:03 +03:00
for hash := range queue {
data, err := srcDb.TrieDB().Node(hash)
if err != nil {
data, err = srcDb.ContractCode(common.Hash{}, hash)
}
if err != nil {
t.Fatalf("failed to retrieve node data for %x", hash)
}
2016-04-15 12:06:57 +03:00
results = append(results, trie.SyncResult{Hash: hash, Data: data})
}
// Feed the retrieved results back and queue new tasks
for _, result := range results {
if err := sched.Process(result); err != nil {
t.Fatalf("failed to process result %v", err)
}
}
batch := dstDb.NewBatch()
if err := sched.Commit(batch); err != nil {
t.Fatalf("failed to commit data: %v", err)
eth/downloader: separate state sync from queue (#14460) * eth/downloader: separate state sync from queue Scheduling of state node downloads hogged the downloader queue lock when new requests were scheduled. This caused timeouts for other requests. With this change, state sync is fully independent of all other downloads and doesn't involve the queue at all. State sync is started and checked on in processContent. This is slightly awkward because processContent doesn't have a select loop. Instead, the queue is closed by an auxiliary goroutine when state sync fails. We tried several alternatives to this but settled on the current approach because it's the least amount of change overall. Handling of the pivot block has changed slightly: the queue previously prevented import of pivot block receipts before the state of the pivot block was available. In this commit, the receipt will be imported before the state. This causes an annoyance where the pivot block is committed as fast block head even when state downloads fail. Stay tuned for more updates in this area ;) * eth/downloader: remove cancelTimeout channel * eth/downloader: retry state requests on timeout * eth/downloader: improve comment * eth/downloader: mark peers idle when state sync is done * eth/downloader: move pivot block splitting to processContent This change also ensures that pivot block receipts aren't imported before the pivot block itself. * eth/downloader: limit state node retries * eth/downloader: improve state node error handling and retry check * eth/downloader: remove maxStateNodeRetries It fails the sync too much. * eth/downloader: remove last use of cancelCh in statesync.go Fixes TestDeliverHeadersHang*Fast and (hopefully) the weird cancellation behaviour at the end of fast sync. * eth/downloader: fix leak in runStateSync * eth/downloader: don't run processFullSyncContent in LightSync mode * eth/downloader: improve comments * eth/downloader: fix vet, megacheck * eth/downloader: remove unrequested tasks anyway * eth/downloader, trie: various polishes around duplicate items This commit explicitly tracks duplicate and unexpected state delieveries done against a trie Sync structure, also adding there to import info logs. The commit moves the db batch used to commit trie changes one level deeper so its flushed after every node insertion. This is needed to avoid a lot of duplicate retrievals caused by inconsistencies between Sync internals and database. A better approach is to track not-yet-written states in trie.Sync and flush on commit, but I'm focuing on correctness first now. The commit fixes a regression around pivot block fail count. The counter previously was reset to 1 if and only if a sync cycle progressed (inserted at least 1 entry to the database). The current code reset it already if a node was delivered, which is not stong enough, because unless it ends up written to disk, an attacker can just loop and attack ad infinitum. The commit also fixes a regression around state deliveries and timeouts. The old downloader tracked if a delivery is stale (none of the deliveries were requestedt), in which case it didn't mark the node idle and did not send further requests, since it signals a past timeout. The current code did mark it idle even on stale deliveries, which eventually caused two requests to be in flight at the same time, making the deliveries always stale and mass duplicating retrievals between multiple peers. * eth/downloader: fix state request leak This commit fixes the hang seen sometimes while doing the state sync. The cause of the hang was a rare combination of events: request state data from peer, peer drops and reconnects almost immediately. This caused a new download task to be assigned to the peer, overwriting the old one still waiting for a timeout, which in turned leaked the requests out, never to be retried. The fix is to ensure that a task assignment moves any pending one back into the retry queue. The commit also fixes a regression with peer dropping due to stalls. The current code considered a peer stalling if they timed out delivering 1 item. However, the downloader never requests only one, the minimum is 2 (attempt to fine tune estimated latency/bandwidth). The fix is simply to drop if a timeout is detected at 2 items. Apart from the above bugfixes, the commit contains some code polishes I made while debugging the hang. * core, eth, trie: support batched trie sync db writes * trie: rename SyncMemCache to syncMemBatch
2017-06-22 15:26:03 +03:00
}
batch.Write()
queue = make(map[common.Hash]struct{})
nodes, _, codes = sched.Missing(count)
for _, hash := range append(nodes, codes...) {
queue[hash] = struct{}{}
}
}
// Cross check that the two states are in sync
checkStateAccounts(t, dstDb, srcRoot, srcAccounts)
}
// Tests that the trie scheduler can correctly reconstruct the state even if only
// partial results are returned (Even those randomly), others sent only later.
func TestIterativeRandomDelayedStateSync(t *testing.T) {
// Create a random state to copy
srcDb, srcRoot, srcAccounts := makeTestState()
// Create a destination state and sync with the scheduler
dstDb := rawdb.NewMemoryDatabase()
core, eth: faster snapshot generation (#22504) * eth/protocols: persist received state segments * core: initial implementation * core/state/snapshot: add tests * core, eth: updates * eth/protocols/snapshot: count flat state size * core/state: add metrics * core/state/snapshot: skip unnecessary deletion * core/state/snapshot: rename * core/state/snapshot: use the global batch * core/state/snapshot: add logs and fix wiping * core/state/snapshot: fix * core/state/snapshot: save generation progress even if the batch is empty * core/state/snapshot: fixes * core/state/snapshot: fix initial account range length * core/state/snapshot: fix initial account range * eth/protocols/snap: store flat states during the healing * eth/protocols/snap: print logs * core/state/snapshot: refactor (#4) * core/state/snapshot: refactor * core/state/snapshot: tiny fix and polish Co-authored-by: rjl493456442 <garyrong0905@gmail.com> * core, eth: fixes * core, eth: fix healing writer * core, trie, eth: fix paths * eth/protocols/snap: fix encoding * eth, core: add debug log * core/state/generate: release iterator asap (#5) core/state/snapshot: less copy core/state/snapshot: revert split loop core/state/snapshot: handle storage becoming empty, improve test robustness core/state: test modified codehash core/state/snapshot: polish * core/state/snapshot: optimize stats counter * core, eth: add metric * core/state/snapshot: update comments * core/state/snapshot: improve tests * core/state/snapshot: replace secure trie with standard trie * core/state/snapshot: wrap return as the struct * core/state/snapshot: skip wiping correct states * core/state/snapshot: updates * core/state/snapshot: fixes * core/state/snapshot: fix panic due to reference flaw in closure * core/state/snapshot: fix errors in state generation logic + fix log output * core/state/snapshot: remove an error case * core/state/snapshot: fix condition-check for exhausted snap state * core/state/snapshot: use stackTrie for small tries * core/state/snapshot: don't resolve small storage tries in vain * core/state/snapshot: properly clean up storage of deleted accounts * core/state/snapshot: avoid RLP-encoding in some cases + minor nitpicks * core/state/snapshot: fix error (+testcase) * core/state/snapshot: clean up tests a bit * core/state/snapshot: work in progress on better tests * core/state/snapshot: polish code * core/state/snapshot: fix trie iteration abortion trigger * core/state/snapshot: fixes flaws * core/state/snapshot: remove panic * core/state/snapshot: fix abort * core/state/snapshot: more tests (plus failing testcase) * core/state/snapshot: more testcases + fix for failing test * core/state/snapshot: testcase for malformed data * core/state/snapshot: some test nitpicks * core/state/snapshot: improvements to logging * core/state/snapshot: testcase to demo error in abortion * core/state/snapshot: fix abortion * cmd/geth: make verify-state report the root * trie: fix failing test * core/state/snapshot: add timer metrics * core/state/snapshot: fix metrics * core/state/snapshot: udpate tests * eth/protocols/snap: write snapshot account even if code or state is needed * core/state/snapshot: fix diskmore check * core/state/snapshot: review fixes * core/state/snapshot: improve error message * cmd/geth: rename 'error' to 'err' in logs * core/state/snapshot: fix some review concerns * core/state/snapshot, eth/protocols/snap: clear snapshot marker when starting/resuming snap sync * core: add error log * core/state/snapshot: use proper timers for metrics collection * core/state/snapshot: address some review concerns * eth/protocols/snap: improved log message * eth/protocols/snap: fix heal logs to condense infos * core/state/snapshot: wait for generator termination before restarting * core/state/snapshot: revert timers to counters to track total time Co-authored-by: Martin Holst Swende <martin@swende.se> Co-authored-by: Péter Szilágyi <peterke@gmail.com>
2021-04-14 23:23:11 +03:00
sched := NewStateSync(srcRoot, dstDb, trie.NewSyncBloom(1, dstDb), nil)
queue := make(map[common.Hash]struct{})
nodes, _, codes := sched.Missing(0)
for _, hash := range append(nodes, codes...) {
queue[hash] = struct{}{}
}
for len(queue) > 0 {
// Sync only half of the scheduled nodes, even those in random order
results := make([]trie.SyncResult, 0, len(queue)/2+1)
2017-01-06 17:52:03 +03:00
for hash := range queue {
delete(queue, hash)
data, err := srcDb.TrieDB().Node(hash)
if err != nil {
data, err = srcDb.ContractCode(common.Hash{}, hash)
}
if err != nil {
t.Fatalf("failed to retrieve node data for %x", hash)
}
2016-04-15 12:06:57 +03:00
results = append(results, trie.SyncResult{Hash: hash, Data: data})
if len(results) >= cap(results) {
break
}
}
// Feed the retrieved results back and queue new tasks
for _, result := range results {
if err := sched.Process(result); err != nil {
t.Fatalf("failed to process result %v", err)
}
}
batch := dstDb.NewBatch()
if err := sched.Commit(batch); err != nil {
t.Fatalf("failed to commit data: %v", err)
eth/downloader: separate state sync from queue (#14460) * eth/downloader: separate state sync from queue Scheduling of state node downloads hogged the downloader queue lock when new requests were scheduled. This caused timeouts for other requests. With this change, state sync is fully independent of all other downloads and doesn't involve the queue at all. State sync is started and checked on in processContent. This is slightly awkward because processContent doesn't have a select loop. Instead, the queue is closed by an auxiliary goroutine when state sync fails. We tried several alternatives to this but settled on the current approach because it's the least amount of change overall. Handling of the pivot block has changed slightly: the queue previously prevented import of pivot block receipts before the state of the pivot block was available. In this commit, the receipt will be imported before the state. This causes an annoyance where the pivot block is committed as fast block head even when state downloads fail. Stay tuned for more updates in this area ;) * eth/downloader: remove cancelTimeout channel * eth/downloader: retry state requests on timeout * eth/downloader: improve comment * eth/downloader: mark peers idle when state sync is done * eth/downloader: move pivot block splitting to processContent This change also ensures that pivot block receipts aren't imported before the pivot block itself. * eth/downloader: limit state node retries * eth/downloader: improve state node error handling and retry check * eth/downloader: remove maxStateNodeRetries It fails the sync too much. * eth/downloader: remove last use of cancelCh in statesync.go Fixes TestDeliverHeadersHang*Fast and (hopefully) the weird cancellation behaviour at the end of fast sync. * eth/downloader: fix leak in runStateSync * eth/downloader: don't run processFullSyncContent in LightSync mode * eth/downloader: improve comments * eth/downloader: fix vet, megacheck * eth/downloader: remove unrequested tasks anyway * eth/downloader, trie: various polishes around duplicate items This commit explicitly tracks duplicate and unexpected state delieveries done against a trie Sync structure, also adding there to import info logs. The commit moves the db batch used to commit trie changes one level deeper so its flushed after every node insertion. This is needed to avoid a lot of duplicate retrievals caused by inconsistencies between Sync internals and database. A better approach is to track not-yet-written states in trie.Sync and flush on commit, but I'm focuing on correctness first now. The commit fixes a regression around pivot block fail count. The counter previously was reset to 1 if and only if a sync cycle progressed (inserted at least 1 entry to the database). The current code reset it already if a node was delivered, which is not stong enough, because unless it ends up written to disk, an attacker can just loop and attack ad infinitum. The commit also fixes a regression around state deliveries and timeouts. The old downloader tracked if a delivery is stale (none of the deliveries were requestedt), in which case it didn't mark the node idle and did not send further requests, since it signals a past timeout. The current code did mark it idle even on stale deliveries, which eventually caused two requests to be in flight at the same time, making the deliveries always stale and mass duplicating retrievals between multiple peers. * eth/downloader: fix state request leak This commit fixes the hang seen sometimes while doing the state sync. The cause of the hang was a rare combination of events: request state data from peer, peer drops and reconnects almost immediately. This caused a new download task to be assigned to the peer, overwriting the old one still waiting for a timeout, which in turned leaked the requests out, never to be retried. The fix is to ensure that a task assignment moves any pending one back into the retry queue. The commit also fixes a regression with peer dropping due to stalls. The current code considered a peer stalling if they timed out delivering 1 item. However, the downloader never requests only one, the minimum is 2 (attempt to fine tune estimated latency/bandwidth). The fix is simply to drop if a timeout is detected at 2 items. Apart from the above bugfixes, the commit contains some code polishes I made while debugging the hang. * core, eth, trie: support batched trie sync db writes * trie: rename SyncMemCache to syncMemBatch
2017-06-22 15:26:03 +03:00
}
batch.Write()
for _, result := range results {
delete(queue, result.Hash)
}
nodes, _, codes = sched.Missing(0)
for _, hash := range append(nodes, codes...) {
queue[hash] = struct{}{}
}
}
// Cross check that the two states are in sync
checkStateAccounts(t, dstDb, srcRoot, srcAccounts)
}
// Tests that at any point in time during a sync, only complete sub-tries are in
// the database.
func TestIncompleteStateSync(t *testing.T) {
// Create a random state to copy
srcDb, srcRoot, srcAccounts := makeTestState()
// isCodeLookup to save some hashing
var isCode = make(map[common.Hash]struct{})
for _, acc := range srcAccounts {
if len(acc.code) > 0 {
isCode[crypto.Keccak256Hash(acc.code)] = struct{}{}
}
}
isCode[common.BytesToHash(emptyCodeHash)] = struct{}{}
checkTrieConsistency(srcDb.TrieDB().DiskDB().(ethdb.Database), srcRoot)
// Create a destination state and sync with the scheduler
dstDb := rawdb.NewMemoryDatabase()
core, eth: faster snapshot generation (#22504) * eth/protocols: persist received state segments * core: initial implementation * core/state/snapshot: add tests * core, eth: updates * eth/protocols/snapshot: count flat state size * core/state: add metrics * core/state/snapshot: skip unnecessary deletion * core/state/snapshot: rename * core/state/snapshot: use the global batch * core/state/snapshot: add logs and fix wiping * core/state/snapshot: fix * core/state/snapshot: save generation progress even if the batch is empty * core/state/snapshot: fixes * core/state/snapshot: fix initial account range length * core/state/snapshot: fix initial account range * eth/protocols/snap: store flat states during the healing * eth/protocols/snap: print logs * core/state/snapshot: refactor (#4) * core/state/snapshot: refactor * core/state/snapshot: tiny fix and polish Co-authored-by: rjl493456442 <garyrong0905@gmail.com> * core, eth: fixes * core, eth: fix healing writer * core, trie, eth: fix paths * eth/protocols/snap: fix encoding * eth, core: add debug log * core/state/generate: release iterator asap (#5) core/state/snapshot: less copy core/state/snapshot: revert split loop core/state/snapshot: handle storage becoming empty, improve test robustness core/state: test modified codehash core/state/snapshot: polish * core/state/snapshot: optimize stats counter * core, eth: add metric * core/state/snapshot: update comments * core/state/snapshot: improve tests * core/state/snapshot: replace secure trie with standard trie * core/state/snapshot: wrap return as the struct * core/state/snapshot: skip wiping correct states * core/state/snapshot: updates * core/state/snapshot: fixes * core/state/snapshot: fix panic due to reference flaw in closure * core/state/snapshot: fix errors in state generation logic + fix log output * core/state/snapshot: remove an error case * core/state/snapshot: fix condition-check for exhausted snap state * core/state/snapshot: use stackTrie for small tries * core/state/snapshot: don't resolve small storage tries in vain * core/state/snapshot: properly clean up storage of deleted accounts * core/state/snapshot: avoid RLP-encoding in some cases + minor nitpicks * core/state/snapshot: fix error (+testcase) * core/state/snapshot: clean up tests a bit * core/state/snapshot: work in progress on better tests * core/state/snapshot: polish code * core/state/snapshot: fix trie iteration abortion trigger * core/state/snapshot: fixes flaws * core/state/snapshot: remove panic * core/state/snapshot: fix abort * core/state/snapshot: more tests (plus failing testcase) * core/state/snapshot: more testcases + fix for failing test * core/state/snapshot: testcase for malformed data * core/state/snapshot: some test nitpicks * core/state/snapshot: improvements to logging * core/state/snapshot: testcase to demo error in abortion * core/state/snapshot: fix abortion * cmd/geth: make verify-state report the root * trie: fix failing test * core/state/snapshot: add timer metrics * core/state/snapshot: fix metrics * core/state/snapshot: udpate tests * eth/protocols/snap: write snapshot account even if code or state is needed * core/state/snapshot: fix diskmore check * core/state/snapshot: review fixes * core/state/snapshot: improve error message * cmd/geth: rename 'error' to 'err' in logs * core/state/snapshot: fix some review concerns * core/state/snapshot, eth/protocols/snap: clear snapshot marker when starting/resuming snap sync * core: add error log * core/state/snapshot: use proper timers for metrics collection * core/state/snapshot: address some review concerns * eth/protocols/snap: improved log message * eth/protocols/snap: fix heal logs to condense infos * core/state/snapshot: wait for generator termination before restarting * core/state/snapshot: revert timers to counters to track total time Co-authored-by: Martin Holst Swende <martin@swende.se> Co-authored-by: Péter Szilágyi <peterke@gmail.com>
2021-04-14 23:23:11 +03:00
sched := NewStateSync(srcRoot, dstDb, trie.NewSyncBloom(1, dstDb), nil)
var added []common.Hash
nodes, _, codes := sched.Missing(1)
queue := append(append([]common.Hash{}, nodes...), codes...)
for len(queue) > 0 {
// Fetch a batch of state nodes
results := make([]trie.SyncResult, len(queue))
for i, hash := range queue {
data, err := srcDb.TrieDB().Node(hash)
if err != nil {
data, err = srcDb.ContractCode(common.Hash{}, hash)
}
if err != nil {
t.Fatalf("failed to retrieve node data for %x", hash)
}
2016-04-15 12:06:57 +03:00
results[i] = trie.SyncResult{Hash: hash, Data: data}
}
// Process each of the state nodes
for _, result := range results {
if err := sched.Process(result); err != nil {
t.Fatalf("failed to process result %v", err)
}
}
batch := dstDb.NewBatch()
if err := sched.Commit(batch); err != nil {
t.Fatalf("failed to commit data: %v", err)
eth/downloader: separate state sync from queue (#14460) * eth/downloader: separate state sync from queue Scheduling of state node downloads hogged the downloader queue lock when new requests were scheduled. This caused timeouts for other requests. With this change, state sync is fully independent of all other downloads and doesn't involve the queue at all. State sync is started and checked on in processContent. This is slightly awkward because processContent doesn't have a select loop. Instead, the queue is closed by an auxiliary goroutine when state sync fails. We tried several alternatives to this but settled on the current approach because it's the least amount of change overall. Handling of the pivot block has changed slightly: the queue previously prevented import of pivot block receipts before the state of the pivot block was available. In this commit, the receipt will be imported before the state. This causes an annoyance where the pivot block is committed as fast block head even when state downloads fail. Stay tuned for more updates in this area ;) * eth/downloader: remove cancelTimeout channel * eth/downloader: retry state requests on timeout * eth/downloader: improve comment * eth/downloader: mark peers idle when state sync is done * eth/downloader: move pivot block splitting to processContent This change also ensures that pivot block receipts aren't imported before the pivot block itself. * eth/downloader: limit state node retries * eth/downloader: improve state node error handling and retry check * eth/downloader: remove maxStateNodeRetries It fails the sync too much. * eth/downloader: remove last use of cancelCh in statesync.go Fixes TestDeliverHeadersHang*Fast and (hopefully) the weird cancellation behaviour at the end of fast sync. * eth/downloader: fix leak in runStateSync * eth/downloader: don't run processFullSyncContent in LightSync mode * eth/downloader: improve comments * eth/downloader: fix vet, megacheck * eth/downloader: remove unrequested tasks anyway * eth/downloader, trie: various polishes around duplicate items This commit explicitly tracks duplicate and unexpected state delieveries done against a trie Sync structure, also adding there to import info logs. The commit moves the db batch used to commit trie changes one level deeper so its flushed after every node insertion. This is needed to avoid a lot of duplicate retrievals caused by inconsistencies between Sync internals and database. A better approach is to track not-yet-written states in trie.Sync and flush on commit, but I'm focuing on correctness first now. The commit fixes a regression around pivot block fail count. The counter previously was reset to 1 if and only if a sync cycle progressed (inserted at least 1 entry to the database). The current code reset it already if a node was delivered, which is not stong enough, because unless it ends up written to disk, an attacker can just loop and attack ad infinitum. The commit also fixes a regression around state deliveries and timeouts. The old downloader tracked if a delivery is stale (none of the deliveries were requestedt), in which case it didn't mark the node idle and did not send further requests, since it signals a past timeout. The current code did mark it idle even on stale deliveries, which eventually caused two requests to be in flight at the same time, making the deliveries always stale and mass duplicating retrievals between multiple peers. * eth/downloader: fix state request leak This commit fixes the hang seen sometimes while doing the state sync. The cause of the hang was a rare combination of events: request state data from peer, peer drops and reconnects almost immediately. This caused a new download task to be assigned to the peer, overwriting the old one still waiting for a timeout, which in turned leaked the requests out, never to be retried. The fix is to ensure that a task assignment moves any pending one back into the retry queue. The commit also fixes a regression with peer dropping due to stalls. The current code considered a peer stalling if they timed out delivering 1 item. However, the downloader never requests only one, the minimum is 2 (attempt to fine tune estimated latency/bandwidth). The fix is simply to drop if a timeout is detected at 2 items. Apart from the above bugfixes, the commit contains some code polishes I made while debugging the hang. * core, eth, trie: support batched trie sync db writes * trie: rename SyncMemCache to syncMemBatch
2017-06-22 15:26:03 +03:00
}
batch.Write()
for _, result := range results {
added = append(added, result.Hash)
// Check that all known sub-tries added so far are complete or missing entirely.
if _, ok := isCode[result.Hash]; ok {
continue
}
// Can't use checkStateConsistency here because subtrie keys may have odd
// length and crash in LeafKey.
if err := checkTrieConsistency(dstDb, result.Hash); err != nil {
t.Fatalf("state inconsistent: %v", err)
}
}
// Fetch the next batch to retrieve
nodes, _, codes = sched.Missing(1)
queue = append(append(queue[:0], nodes...), codes...)
}
// Sanity check that removing any node from the database is detected
for _, node := range added[1:] {
var (
key = node.Bytes()
_, code = isCode[node]
val []byte
)
if code {
val = rawdb.ReadCode(dstDb, node)
rawdb.DeleteCode(dstDb, node)
} else {
val = rawdb.ReadTrieNode(dstDb, node)
rawdb.DeleteTrieNode(dstDb, node)
}
if err := checkStateConsistency(dstDb, added[0]); err == nil {
t.Fatalf("trie inconsistency not caught, missing: %x", key)
}
if code {
rawdb.WriteCode(dstDb, node, val)
} else {
rawdb.WriteTrieNode(dstDb, node, val)
}
}
}