Files
lotus/conformance/driver.go
Aarsh Shah dcc903c65d feat: a new ChainIndexer to index tipsets, messages and events (#12421)
* chain index complete for msgs and txns

* dont need observer changes for now

* changes

* fix tests

* fix tests

* use th right context

* index empty tipsets correctly

* implement automated backfilling

* add event indexing and remove all old indices

* fix test

* revert deployment test changes

* revert test changes and better error handling for eth tx index lookups

* fix sql statments naming convention

* address review for Index GC

* more changes as per review

* changes as per review

* fix config

* mark events as reverted during reconciliation

* better reconciliation; pens down and code complete; also reconcile events

* fix tests

* improve config and docs

* improve docs and error handling

* improve read logic

* improve docs

* better logging and handle ennable event storage

* improve logs and index init proc

* better logging

* fix bugs based on calibnet testing

* create sqliite Indices

* gc should be based on epochs

* fix event query

* foreign keys should be enabled on the DB

* reverted tipsets should be removed as part of GC

* release read lock

* make it easy to backfill an empty index using reconciliation

* better docs for reconciliation

* fix conflicts with master

* Apply suggestions from code review

Co-authored-by: Rod Vagg <rod@vagg.org>

* fix go mod

* fix formatting

* revert config changes

* address changes in observer

* remove top level chainindex package

* changes as per review

* changes as per review

* changes as per review

* handle index with reverted tipsets during reconciliation

* changes as per review

* fix type of max reconcile epoch

* changes to reconciliation as per review

* log ipld error

* better logging of progress

* disable chain indexer hydrate from snapshot based on config

* always populate index

* make config easy to reason about

* fix config

* fix messaging

* revert config changes

* Apply suggestions from code review

Co-authored-by: Rod Vagg <rod@vagg.org>

* changes as per review

* make error messages homogenous

* fix indentation

* changes as per review

* feat: recompute tipset to generate missing events if event indexing is enabled (#12463)

* auto repair events

* make jen

* fix leaky abstraction

* better docs for gc retention epoch

* imrpove DB handling (#12485)

* fix conflict

* fix lite node config for indexer

* exclude reverted events from eth get logs if client queries by epoch

* Simply addressing for event lookups in the index.

simply addressing for event lookups

* Apply suggestions from code review

Co-authored-by: Rod Vagg <rod@vagg.org>

* Apply suggestions from code review

Co-authored-by: Rod Vagg <rod@vagg.org>

* fix tests

* Apply suggestions from code review

Co-authored-by: Rod Vagg <rod@vagg.org>

* feat: migration("re-indexing"), backfilling and diasgnostics tooling for the `ChainIndexer` (#12450)

* fix conflicts with chain indexer

* feat: chain indexer todos [skip changelog] (#12462)

* feat: finish todos of validation api

* feat: add indexed data verification with chain store

* feat: address comments and finish TODO

* fix: build issue

* address comments

* fix: ci issue

* Apply suggestions from code review

Co-authored-by: Rod Vagg <rod@vagg.org>

* changes to Index Validation API based on Rodds first review

* build chain indexer API

* improve error handling

* feat: lotus-shed tooling for chain indexer (#12474)

* feat: add lotus-shed command for backfilling chain indexer

* feat: add lotus-shed command for inspecting the chain indexer

* feat: use single lotus-shed command to inspect and backfill

* fix: remove the unused queries

* small changes

* add change log

* backfilling improvements and fixes

* finish chain index validation and backfill tooling

* user documentation for the

* validate from epoch

* Apply suggestions from code review

Suggestions from Steve's read of the user doc.

Co-authored-by: Steve Loeppky <biglep@filoz.org>

* changes to user doc as per review

* Apply suggestions from code review

Co-authored-by: Steve Loeppky <biglep@filoz.org>

* changes to user doc as per review

* Apply suggestions from code review

Co-authored-by: Steve Loeppky <biglep@filoz.org>

* changes as per review

* feat: add event entries count in validation API (#12506)

* feat: add event entry count in validation API

* address comments

* use sqllite defaults (#12504)

* Apply suggestions from code review

Co-authored-by: Steve Loeppky <biglep@filoz.org>

* write chain index to a different dir

* Apply suggestions from code review

Co-authored-by: Steve Loeppky <biglep@filoz.org>

* fix conflicts

* UX improvements to backfilling

* feat: tests for the chain indexer (#12521)

* ddl tests

* tests for the chain indexer

* finish unit tests for chain indexer

* fix formatting

* cleanup reverted tipsets to avoid db bloat

* fix logging

* test for filter by address

* test gc cascade delete

* fix db locked error during backfilling

* fix var name

* increase db locked timeout

* fix db locked issue

* reduce db lock timeout

* no lock in gc

* reconcile does not need lock

* improved error handling

* Update chain-indexing-overview-for-rpc-providers.md

Doc updates based on @jennijuju feedack.

* Update chain-indexing-overview-for-rpc-providers.MD

Fixes after reviewing 33c1ca1831

* better metrics for backfilling

* Update chain/index/chain-indexing-overview-for-rpc-providers.MD

Co-authored-by: Rod Vagg <rod@vagg.org>

* Update chain/index/chain-indexing-overview-for-rpc-providers.MD

Co-authored-by: Rod Vagg <rod@vagg.org>

* Update chain/index/chain-indexing-overview-for-rpc-providers.MD

Co-authored-by: Rod Vagg <rod@vagg.org>

* Update chain/index/chain-indexing-overview-for-rpc-providers.MD

Co-authored-by: Rod Vagg <rod@vagg.org>

* Update chain/index/chain-indexing-overview-for-rpc-providers.MD

Co-authored-by: Rod Vagg <rod@vagg.org>

* Update chain/index/chain-indexing-overview-for-rpc-providers.MD

Co-authored-by: Rod Vagg <rod@vagg.org>

* Update chain/index/chain-indexing-overview-for-rpc-providers.MD

Co-authored-by: Rod Vagg <rod@vagg.org>

* tests for changes to event addressing

* Apply suggestions from code review

Co-authored-by: Rod Vagg <rod@vagg.org>

* changes as per review -> round 1

* Apply suggestions from code review

Co-authored-by: Rod Vagg <rod@vagg.org>

* Apply suggestions from code review

Co-authored-by: Rod Vagg <rod@vagg.org>

* log tipset key cid

* Apply suggestions from code review

Co-authored-by: Rod Vagg <rod@vagg.org>

* Apply suggestions from code review

Co-authored-by: Rod Vagg <rod@vagg.org>

* fix docs

* Apply suggestions from code review

Co-authored-by: Rod Vagg <rod@vagg.org>

* fix tests

* fix tests

* make jen

* fix conflicts

---------

Co-authored-by: Aryan Tikarya <aryan.tikarya@dojima.network>
Co-authored-by: Rod Vagg <rod@vagg.org>
Co-authored-by: Steve Loeppky <biglep@filoz.org>

* fix lint

* Apply suggestions from code review

Co-authored-by: Rod Vagg <rod@vagg.org>

* Apply suggestions from code review

Co-authored-by: Rod Vagg <rod@vagg.org>

* remove reverted flag from RPC

* Apply suggestions from code review

Co-authored-by: Rod Vagg <rod@vagg.org>

* fix testing of events and dummy chain store

* remove lotus shed commands for old Indices

* change type of event counts to uint64

* only recompute events if theyre not found

* short-circuit empty events path for older tipsets

* chain indexer must be enabled if ETH RPC is enabled

* change name of message_id column to id in tipset_message table

* only expose SetRecomputeTipSetStateFunc

* dont block on head indexing for reading messages

* document why we're only checking for missing events for a single tipset

* document when we query for reverted events

* simplify event collection

* Apply suggestions from code review

Co-authored-by: Rod Vagg <rod@vagg.org>

* fix test

* change event_id to id in the event table

* change head indexed timeout

* remove deprecated config options

* fail ETH RPC calls if ChainIndexer is disabled

* fix docs

* remove the tipset key cid func from lotus shed

* address review comments

* Apply suggestions from code review

Co-authored-by: Rod Vagg <rod@vagg.org>

* chore(events): remove unnecessary DisableRealTimeFilterAPI (#12610)

* feat(cli): add --quiet to chainindex validate-backfill + cleanups (#12611)

* fix tests

* Apply suggestions from code review

Co-authored-by: Rod Vagg <rod@vagg.org>

* error type for disabled chainindexer

* fix(chainindex): recompute tipset when we find no receipts

* fix(chainindexer): backfilling should halt when chain state data is missing and not backfill parents (#12619)

* fix backfilling UX

* Update chain/index/api.go

Co-authored-by: Rod Vagg <rod@vagg.org>

* address review

---------

Co-authored-by: Rod Vagg <rod@vagg.org>

* reduce log noise

* make jen

* make jen

* docs: finishing chain-indexer-overview-for-operators.md (#12600)

* Followup to PR #12450 for doc updates

This is being used to resolve the unresolved items in https://github.com/filecoin-project/lotus/pull/12450 since that PR is unwieldly at this point.

* Incorporated some items and added TODOs based on unresolved items from https://github.com/filecoin-project/lotus/pull/12450

* Incorporating more feedback

* Pointing to issue to learn about benefits

* Formatting fixes

* Apply most of the suggestions from @rvagg code review

Co-authored-by: Rod Vagg <rod@vagg.org>

* Incorporating feedback from https://github.com/filecoin-project/lotus/pull/12600#discussion_r1802519453

* Addressing https://github.com/filecoin-project/lotus/pull/12600#discussion_r1802540042 and more

* Moved chain-indexer docs to documentation
Renamed
Added ToC

We can move to lotus-docs later

* Update documentation/en/chain-indexer-overview-for-operators.md

Co-authored-by: Rod Vagg <rod@vagg.org>

* Update documentation/en/chain-indexer-overview-for-operators.md

Co-authored-by: Rod Vagg <rod@vagg.org>

* Added upgrade path when importing chain state from a snapshot.

* Typo fixes

* Update documentation/en/chain-indexer-overview-for-operators.md

Co-authored-by: Rod Vagg <rod@vagg.org>

* chore(doc): "regular checks" section for chainindexer docs (#12612)

* Apply suggestions from @rvagg code review

Co-authored-by: Rod Vagg <rod@vagg.org>

* Incorporating @aarshkshah1992 feedback

* Update documentation/en/chain-indexer-overview-for-operators.md

Co-authored-by: Rod Vagg <rod@vagg.org>

---------

Co-authored-by: Rod Vagg <rod@vagg.org>
Co-authored-by: Aarsh Shah <aarshkshah1992@gmail.com>

* remove go mod replace

* remove unnecessary changes from CHANGELOG

* fix test

* compare events AMT root (#12632)

* fix(chainindex): retry transaction if database connection is lost (#12657)

* retry database lost connection

* log context cancellation

* address review

* fix gateway itest: no chainindexer for lite nodes

* fix changelog

---------

Co-authored-by: Rod Vagg <rod@vagg.org>
Co-authored-by: Aryan Tikarya <aryan.tikarya@dojima.network>
Co-authored-by: Steve Loeppky <biglep@filoz.org>
2024-10-31 09:58:19 +00:00

362 lines
12 KiB
Go

package conformance
import (
"context"
gobig "math/big"
"os"
"github.com/ipfs/go-cid"
ds "github.com/ipfs/go-datastore"
cbor "github.com/ipfs/go-ipld-cbor"
"github.com/filecoin-project/go-address"
"github.com/filecoin-project/go-state-types/abi"
actorstypes "github.com/filecoin-project/go-state-types/actors"
"github.com/filecoin-project/go-state-types/big"
"github.com/filecoin-project/go-state-types/crypto"
"github.com/filecoin-project/go-state-types/network"
rtt "github.com/filecoin-project/go-state-types/rt"
"github.com/filecoin-project/test-vectors/schema"
"github.com/filecoin-project/lotus/blockstore"
"github.com/filecoin-project/lotus/chain/actors/builtin"
"github.com/filecoin-project/lotus/chain/consensus"
"github.com/filecoin-project/lotus/chain/consensus/filcns"
proofsffi "github.com/filecoin-project/lotus/chain/proofs/ffi"
"github.com/filecoin-project/lotus/chain/rand"
"github.com/filecoin-project/lotus/chain/state"
"github.com/filecoin-project/lotus/chain/stmgr"
"github.com/filecoin-project/lotus/chain/store"
"github.com/filecoin-project/lotus/chain/types"
"github.com/filecoin-project/lotus/chain/vm"
"github.com/filecoin-project/lotus/conformance/chaos"
_ "github.com/filecoin-project/lotus/lib/sigs/bls" // enable bls signatures
_ "github.com/filecoin-project/lotus/lib/sigs/delegated"
_ "github.com/filecoin-project/lotus/lib/sigs/secp" // enable secp signatures
)
var (
// DefaultCirculatingSupply is the fallback circulating supply returned by
// the driver's CircSupplyCalculator function, used if the vector specifies
// no circulating supply.
DefaultCirculatingSupply = types.TotalFilecoinInt
// DefaultBaseFee to use in the VM, if one is not supplied in the vector.
DefaultBaseFee = abi.NewTokenAmount(100)
)
type Driver struct {
ctx context.Context
selector schema.Selector
vmFlush bool
}
type DriverOpts struct {
// DisableVMFlush, when true, avoids calling VM.Flush(), forces a blockstore
// recursive copy, from the temporary buffer blockstore, to the real
// system's blockstore. Disabling VM flushing is useful when extracting test
// vectors and trimming state, as we don't want to force an accidental
// deep copy of the state tree.
//
// Disabling VM flushing almost always should go hand-in-hand with
// LOTUS_DISABLE_VM_BUF=iknowitsabadidea. That way, state tree writes are
// immediately committed to the blockstore.
DisableVMFlush bool
}
func NewDriver(ctx context.Context, selector schema.Selector, opts DriverOpts) *Driver {
return &Driver{ctx: ctx, selector: selector, vmFlush: !opts.DisableVMFlush}
}
type ExecuteTipsetResult struct {
ReceiptsRoot cid.Cid
PostStateRoot cid.Cid
// AppliedMessages stores the messages that were applied, in the order they
// were applied. It includes implicit messages (cron, rewards).
AppliedMessages []*types.Message
// AppliedResults stores the results of AppliedMessages, in the same order.
AppliedResults []*vm.ApplyRet
// PostBaseFee returns the basefee after applying this tipset.
PostBaseFee abi.TokenAmount
}
type ExecuteTipsetParams struct {
Preroot cid.Cid
// ParentEpoch is the last epoch in which an actual tipset was processed. This
// is used by Lotus for null block counting and cron firing.
ParentEpoch abi.ChainEpoch
Tipset *schema.Tipset
ExecEpoch abi.ChainEpoch
// Rand is an optional rand.Rand implementation to use. If nil, the driver
// will use a rand.Rand that returns a fixed value for all calls.
Rand rand.Rand
// BaseFee if not nil or zero, will override the basefee of the tipset.
BaseFee abi.TokenAmount
}
// ExecuteTipset executes the supplied tipset on top of the state represented
// by the preroot CID.
//
// This method returns the receipts root, the poststate root, and the VM
// message results. The latter _include_ implicit messages, such as cron ticks
// and reward withdrawal per miner.
func (d *Driver) ExecuteTipset(bs blockstore.Blockstore, ds ds.Batching, params ExecuteTipsetParams) (*ExecuteTipsetResult, error) {
var (
tipset = params.Tipset
syscalls = vm.Syscalls(proofsffi.ProofVerifier)
cs = store.NewChainStore(bs, bs, ds, filcns.Weight, nil)
tse = consensus.NewTipSetExecutor(filcns.RewardFunc)
sm, err = stmgr.NewStateManager(cs, tse, syscalls, filcns.DefaultUpgradeSchedule(), nil, ds, nil)
)
if err != nil {
return nil, err
}
if params.Rand == nil {
params.Rand = NewFixedRand()
}
if params.BaseFee.NilOrZero() {
params.BaseFee = abi.NewTokenAmount(tipset.BaseFee.Int64())
}
defer cs.Close() //nolint:errcheck
blocks := make([]consensus.FilecoinBlockMessages, 0, len(tipset.Blocks))
for _, b := range tipset.Blocks {
sb := store.BlockMessages{
Miner: b.MinerAddr,
}
for _, m := range b.Messages {
msg, err := types.DecodeMessage(m)
if err != nil {
return nil, err
}
switch msg.From.Protocol() {
case address.SECP256K1:
sb.SecpkMessages = append(sb.SecpkMessages, toChainMsg(msg))
case address.BLS:
sb.BlsMessages = append(sb.BlsMessages, toChainMsg(msg))
default:
// sneak in messages originating from other addresses as both kinds.
// these should fail, as they are actually invalid senders.
sb.SecpkMessages = append(sb.SecpkMessages, msg)
sb.BlsMessages = append(sb.BlsMessages, msg)
}
}
blocks = append(blocks, consensus.FilecoinBlockMessages{
BlockMessages: sb,
WinCount: b.WinCount,
})
}
recordOutputs := &outputRecorder{
messages: []*types.Message{},
results: []*vm.ApplyRet{},
}
sm.SetVMConstructor(func(ctx context.Context, vmopt *vm.VMOpts) (vm.Interface, error) {
vmopt.CircSupplyCalc = func(context.Context, abi.ChainEpoch, *state.StateTree) (abi.TokenAmount, error) {
return big.Zero(), nil
}
return vm.NewVM(ctx, vmopt)
})
postcid, receiptsroot, err := tse.ApplyBlocks(context.Background(),
sm,
params.ParentEpoch,
params.Preroot,
blocks,
params.ExecEpoch,
params.Rand,
recordOutputs,
true,
params.BaseFee,
nil,
)
if err != nil {
return nil, err
}
ret := &ExecuteTipsetResult{
ReceiptsRoot: receiptsroot,
PostStateRoot: postcid,
AppliedMessages: recordOutputs.messages,
AppliedResults: recordOutputs.results,
}
return ret, nil
}
type ExecuteMessageParams struct {
Preroot cid.Cid
Epoch abi.ChainEpoch
Timestamp uint64
Message *types.Message
CircSupply abi.TokenAmount
BaseFee abi.TokenAmount
NetworkVersion network.Version
// Rand is an optional rand.Rand implementation to use. If nil, the driver
// will use a rand.Rand that returns a fixed value for all calls.
Rand rand.Rand
// Lookback is the LookbackStateGetter; returns the state tree at a given epoch.
Lookback vm.LookbackStateGetter
// TipSetGetter returns the tipset key at any given epoch.
TipSetGetter vm.TipSetGetter
}
// ExecuteMessage executes a conformance test vector message in a temporary VM.
func (d *Driver) ExecuteMessage(bs blockstore.Blockstore, params ExecuteMessageParams) (*vm.ApplyRet, cid.Cid, error) {
if !d.vmFlush {
// do not flush the VM, just the state tree; this should be used with
// LOTUS_DISABLE_VM_BUF enabled, so writes will anyway be visible.
_ = os.Setenv("LOTUS_DISABLE_VM_BUF", "iknowitsabadidea")
}
if params.Rand == nil {
params.Rand = NewFixedRand()
}
if params.TipSetGetter == nil {
// TODO: If/when we start writing conformance tests against the EVM, we'll need to
// actually implement this and (unfortunately) capture any tipsets looked up by
// messages.
params.TipSetGetter = func(context.Context, abi.ChainEpoch) (types.TipSetKey, error) {
return types.EmptyTSK, nil
}
}
if params.Lookback == nil {
// TODO: This lookback state returns the supplied precondition state tree, unconditionally.
// This is obviously not correct, but the lookback state tree is only used to validate the
// worker key when verifying a consensus fault. If the worker key hasn't changed in the
// current finality window, this workaround is enough.
// The correct solutions are documented in https://github.com/filecoin-project/ref-fvm/issues/381,
// but they're much harder to implement, and the tradeoffs aren't clear.
params.Lookback = func(ctx context.Context, epoch abi.ChainEpoch) (*state.StateTree, error) {
cst := cbor.NewCborStore(bs)
return state.LoadStateTree(cst, params.Preroot)
}
}
vmOpts := &vm.VMOpts{
StateBase: params.Preroot,
Epoch: params.Epoch,
Timestamp: params.Timestamp,
Bstore: bs,
Syscalls: vm.Syscalls(proofsffi.ProofVerifier),
CircSupplyCalc: func(_ context.Context, _ abi.ChainEpoch, _ *state.StateTree) (abi.TokenAmount, error) {
return params.CircSupply, nil
},
Rand: params.Rand,
BaseFee: params.BaseFee,
NetworkVersion: params.NetworkVersion,
LookbackState: params.Lookback,
TipSetGetter: params.TipSetGetter,
}
var vmi vm.Interface
if chaosOn, ok := d.selector["chaos_actor"]; ok && chaosOn == "true" {
lvm, err := vm.NewLegacyVM(context.TODO(), vmOpts)
if err != nil {
return nil, cid.Undef, err
}
invoker := consensus.NewActorRegistry()
av, _ := actorstypes.VersionForNetwork(params.NetworkVersion)
registry := builtin.MakeRegistryLegacy([]rtt.VMActor{chaos.Actor{}})
invoker.Register(av, nil, registry)
lvm.SetInvoker(invoker)
vmi = lvm
} else {
if vmOpts.NetworkVersion >= network.Version16 {
fvm, err := vm.NewFVM(context.TODO(), vmOpts)
if err != nil {
return nil, cid.Undef, err
}
vmi = fvm
} else {
lvm, err := vm.NewLegacyVM(context.TODO(), vmOpts)
if err != nil {
return nil, cid.Undef, err
}
invoker := consensus.NewActorRegistry()
lvm.SetInvoker(invoker)
vmi = lvm
}
}
ret, err := vmi.ApplyMessage(d.ctx, toChainMsg(params.Message))
if err != nil {
return nil, cid.Undef, err
}
var root cid.Cid
if lvm, ok := vmi.(*vm.LegacyVM); ok && !d.vmFlush {
root, err = lvm.StateTree().(*state.StateTree).Flush(d.ctx)
} else {
// flush the VM, committing the state tree changes and forcing a
// recursive copy from the temporary blockstore to the real blockstore.
root, err = vmi.Flush(d.ctx)
}
return ret, root, err
}
// toChainMsg injects a synthetic 0-filled signature of the right length to
// messages that originate from secp256k senders, leaving all
// others untouched.
// TODO: generate a signature in the DSL so that it's encoded in
//
// the test vector.
func toChainMsg(msg *types.Message) (ret types.ChainMsg) {
ret = msg
if msg.From.Protocol() == address.SECP256K1 {
ret = &types.SignedMessage{
Message: *msg,
Signature: crypto.Signature{
Type: crypto.SigTypeSecp256k1,
Data: make([]byte, 65),
},
}
}
return ret
}
// BaseFeeOrDefault converts a basefee as passed in a test vector (go *big.Int
// type) to an abi.TokenAmount, or if nil it returns the DefaultBaseFee.
func BaseFeeOrDefault(basefee *gobig.Int) abi.TokenAmount {
if basefee == nil {
return DefaultBaseFee
}
return big.NewFromGo(basefee)
}
// CircSupplyOrDefault converts a circulating supply as passed in a test vector
// (go *big.Int type) to an abi.TokenAmount, or if nil it returns the
// DefaultCirculatingSupply.
func CircSupplyOrDefault(circSupply *gobig.Int) abi.TokenAmount {
if circSupply == nil {
return DefaultCirculatingSupply
}
return big.NewFromGo(circSupply)
}
type outputRecorder struct {
messages []*types.Message
results []*vm.ApplyRet
}
func (o *outputRecorder) MessageApplied(ctx context.Context, ts *types.TipSet, mcid cid.Cid, msg *types.Message, ret *vm.ApplyRet, implicit bool) error {
o.messages = append(o.messages, msg)
o.results = append(o.results, ret)
return nil
}