neoneo-go/cli/server/server.go
Leonard Lyubich e29c33e449 cli/node: Fix deadlock produced by instant RPC service start
If `StartWhenSynchronized` is unset in config, `node` command runs RPC
service instantly. Previously there was a ground for deadlock. Command
started RPC server synchronously. According to server implementation, it
sends all internal failures to the parameterized error channel. Deadlock
occured because main routine didn't scan the channel.

Run `rpcsrv.Server.Start` in a separate go-routine in `startServer`.
This prevents potential deadlock caused by writing into unread channel.

Fixes #2896.

Signed-off-by: Leonard Lyubich <leonard@morphbits.io>
2023-04-17 10:42:50 +04:00

690 lines
20 KiB
Go

package server
import (
"context"
"errors"
"fmt"
"os"
"os/signal"
"syscall"
"time"
"github.com/nspcc-dev/neo-go/cli/cmdargs"
"github.com/nspcc-dev/neo-go/cli/options"
"github.com/nspcc-dev/neo-go/pkg/config"
"github.com/nspcc-dev/neo-go/pkg/config/netmode"
"github.com/nspcc-dev/neo-go/pkg/consensus"
"github.com/nspcc-dev/neo-go/pkg/core"
"github.com/nspcc-dev/neo-go/pkg/core/block"
"github.com/nspcc-dev/neo-go/pkg/core/chaindump"
corestate "github.com/nspcc-dev/neo-go/pkg/core/stateroot"
"github.com/nspcc-dev/neo-go/pkg/core/storage"
"github.com/nspcc-dev/neo-go/pkg/core/transaction"
"github.com/nspcc-dev/neo-go/pkg/io"
"github.com/nspcc-dev/neo-go/pkg/network"
"github.com/nspcc-dev/neo-go/pkg/services/metrics"
"github.com/nspcc-dev/neo-go/pkg/services/notary"
"github.com/nspcc-dev/neo-go/pkg/services/oracle"
"github.com/nspcc-dev/neo-go/pkg/services/rpcsrv"
"github.com/nspcc-dev/neo-go/pkg/services/stateroot"
"github.com/urfave/cli"
"go.uber.org/zap"
"go.uber.org/zap/zapcore"
)
// NewCommands returns 'node' command.
func NewCommands() []cli.Command {
cfgFlags := []cli.Flag{options.Config}
cfgFlags = append(cfgFlags, options.Network...)
var cfgWithCountFlags = make([]cli.Flag, len(cfgFlags))
copy(cfgWithCountFlags, cfgFlags)
cfgFlags = append(cfgFlags, options.Debug)
cfgWithCountFlags = append(cfgWithCountFlags,
cli.UintFlag{
Name: "count, c",
Usage: "number of blocks to be processed (default or 0: all chain)",
},
)
var cfgCountOutFlags = make([]cli.Flag, len(cfgWithCountFlags))
copy(cfgCountOutFlags, cfgWithCountFlags)
cfgCountOutFlags = append(cfgCountOutFlags,
cli.UintFlag{
Name: "start, s",
Usage: "block number to start from (default: 0)",
},
cli.StringFlag{
Name: "out, o",
Usage: "Output file (stdout if not given)",
},
)
var cfgCountInFlags = make([]cli.Flag, len(cfgWithCountFlags))
copy(cfgCountInFlags, cfgWithCountFlags)
cfgCountInFlags = append(cfgCountInFlags,
cli.StringFlag{
Name: "in, i",
Usage: "Input file (stdin if not given)",
},
cli.StringFlag{
Name: "dump",
Usage: "directory for storing JSON dumps",
},
cli.BoolFlag{
Name: "incremental, n",
Usage: "use if dump is incremental",
},
)
var cfgHeightFlags = make([]cli.Flag, len(cfgFlags)+1)
copy(cfgHeightFlags, cfgFlags)
cfgHeightFlags[len(cfgHeightFlags)-1] = cli.UintFlag{
Name: "height",
Usage: "Height of the state to reset DB to",
Required: true,
}
return []cli.Command{
{
Name: "node",
Usage: "start a NeoGo node",
UsageText: "neo-go node [--config-path path] [-d] [-p/-m/-t]",
Action: startServer,
Flags: cfgFlags,
},
{
Name: "db",
Usage: "database manipulations",
Subcommands: []cli.Command{
{
Name: "dump",
Usage: "dump blocks (starting with block #1) to the file",
UsageText: "neo-go db dump -o file [-s start] [-c count] [--config-path path] [-p/-m/-t]",
Action: dumpDB,
Flags: cfgCountOutFlags,
},
{
Name: "restore",
Usage: "restore blocks from the file",
UsageText: "neo-go db restore -i file [--dump] [-n] [-c count] [--config-path path] [-p/-m/-t]",
Action: restoreDB,
Flags: cfgCountInFlags,
},
{
Name: "reset",
Usage: "reset database to the previous state",
UsageText: "neo-go db reset --height height [--config-path path] [-p/-m/-t]",
Action: resetDB,
Flags: cfgHeightFlags,
},
},
},
}
}
func newGraceContext() context.Context {
ctx, cancel := context.WithCancel(context.Background())
stop := make(chan os.Signal, 1)
signal.Notify(stop, os.Interrupt)
signal.Notify(stop, syscall.SIGTERM)
go func() {
<-stop
cancel()
}()
return ctx
}
func initBCWithMetrics(cfg config.Config, log *zap.Logger) (*core.Blockchain, *metrics.Service, *metrics.Service, error) {
chain, _, err := initBlockChain(cfg, log)
if err != nil {
return nil, nil, nil, cli.NewExitError(err, 1)
}
configureAddresses(&cfg.ApplicationConfiguration)
prometheus := metrics.NewPrometheusService(cfg.ApplicationConfiguration.Prometheus, log)
pprof := metrics.NewPprofService(cfg.ApplicationConfiguration.Pprof, log)
go chain.Run()
err = prometheus.Start()
if err != nil {
return nil, nil, nil, cli.NewExitError(fmt.Errorf("failed to start Prometheus service: %w", err), 1)
}
err = pprof.Start()
if err != nil {
return nil, nil, nil, cli.NewExitError(fmt.Errorf("failed to start Pprof service: %w", err), 1)
}
return chain, prometheus, pprof, nil
}
func dumpDB(ctx *cli.Context) error {
if err := cmdargs.EnsureNone(ctx); err != nil {
return err
}
cfg, err := options.GetConfigFromContext(ctx)
if err != nil {
return cli.NewExitError(err, 1)
}
log, _, logCloser, err := options.HandleLoggingParams(ctx.Bool("debug"), cfg.ApplicationConfiguration)
if err != nil {
return cli.NewExitError(err, 1)
}
if logCloser != nil {
defer func() { _ = logCloser() }()
}
count := uint32(ctx.Uint("count"))
start := uint32(ctx.Uint("start"))
var outStream = os.Stdout
if out := ctx.String("out"); out != "" {
outStream, err = os.Create(out)
if err != nil {
return cli.NewExitError(err, 1)
}
}
defer outStream.Close()
writer := io.NewBinWriterFromIO(outStream)
chain, prometheus, pprof, err := initBCWithMetrics(cfg, log)
if err != nil {
return err
}
defer func() {
pprof.ShutDown()
prometheus.ShutDown()
chain.Close()
}()
chainCount := chain.BlockHeight() + 1
if start+count > chainCount {
return cli.NewExitError(fmt.Errorf("chain is not that high (%d) to dump %d blocks starting from %d", chainCount-1, count, start), 1)
}
if count == 0 {
count = chainCount - start
}
writer.WriteU32LE(count)
err = chaindump.Dump(chain, writer, start, count)
if err != nil {
return cli.NewExitError(err.Error(), 1)
}
return nil
}
func restoreDB(ctx *cli.Context) error {
if err := cmdargs.EnsureNone(ctx); err != nil {
return err
}
cfg, err := options.GetConfigFromContext(ctx)
if err != nil {
return err
}
log, _, logCloser, err := options.HandleLoggingParams(ctx.Bool("debug"), cfg.ApplicationConfiguration)
if err != nil {
return cli.NewExitError(err, 1)
}
if logCloser != nil {
defer func() { _ = logCloser() }()
}
count := uint32(ctx.Uint("count"))
var inStream = os.Stdin
if in := ctx.String("in"); in != "" {
inStream, err = os.Open(in)
if err != nil {
return cli.NewExitError(err, 1)
}
}
defer inStream.Close()
reader := io.NewBinReaderFromIO(inStream)
dumpDir := ctx.String("dump")
if dumpDir != "" {
cfg.ApplicationConfiguration.SaveStorageBatch = true
}
chain, prometheus, pprof, err := initBCWithMetrics(cfg, log)
if err != nil {
return err
}
defer func() {
pprof.ShutDown()
prometheus.ShutDown()
chain.Close()
}()
var start uint32
if ctx.Bool("incremental") {
start = reader.ReadU32LE()
if chain.BlockHeight()+1 < start {
return cli.NewExitError(fmt.Errorf("expected height: %d, dump starts at %d",
chain.BlockHeight()+1, start), 1)
}
}
var skip uint32
if chain.BlockHeight() != 0 {
skip = chain.BlockHeight() + 1 - start
}
var allBlocks = reader.ReadU32LE()
if reader.Err != nil {
return cli.NewExitError(err, 1)
}
if skip+count > allBlocks {
return cli.NewExitError(fmt.Errorf("input file has only %d blocks, can't read %d starting from %d", allBlocks, count, skip), 1)
}
if count == 0 {
count = allBlocks - skip
}
log.Info("initialize restore",
zap.Uint32("start", start),
zap.Uint32("height", chain.BlockHeight()),
zap.Uint32("skip", skip),
zap.Uint32("count", count))
gctx := newGraceContext()
var lastIndex uint32
dump := newDump()
defer func() {
_ = dump.tryPersist(dumpDir, lastIndex)
}()
var f = func(b *block.Block) error {
select {
case <-gctx.Done():
return gctx.Err()
default:
return nil
}
}
if dumpDir != "" {
f = func(b *block.Block) error {
select {
case <-gctx.Done():
return gctx.Err()
default:
}
batch := chain.LastBatch()
// The genesis block may already be persisted, so LastBatch() will return nil.
if batch == nil && b.Index == 0 {
return nil
}
dump.add(b.Index, batch)
lastIndex = b.Index
if b.Index%1000 == 0 {
if err := dump.tryPersist(dumpDir, b.Index); err != nil {
return fmt.Errorf("can't dump storage to file: %w", err)
}
}
return nil
}
}
err = chaindump.Restore(chain, reader, skip, count, f)
if err != nil {
return cli.NewExitError(err, 1)
}
return nil
}
func resetDB(ctx *cli.Context) error {
if err := cmdargs.EnsureNone(ctx); err != nil {
return err
}
cfg, err := options.GetConfigFromContext(ctx)
if err != nil {
return cli.NewExitError(err, 1)
}
h := uint32(ctx.Uint("height"))
log, _, logCloser, err := options.HandleLoggingParams(ctx.Bool("debug"), cfg.ApplicationConfiguration)
if err != nil {
return cli.NewExitError(err, 1)
}
if logCloser != nil {
defer func() { _ = logCloser() }()
}
chain, store, err := initBlockChain(cfg, log)
if err != nil {
return cli.NewExitError(fmt.Errorf("failed to create Blockchain instance: %w", err), 1)
}
err = chain.Reset(h)
if err != nil {
return cli.NewExitError(fmt.Errorf("failed to reset chain state to height %d: %w", h, err), 1)
}
err = store.Close()
if err != nil {
return cli.NewExitError(fmt.Errorf("failed to close the DB: %w", err), 1)
}
return nil
}
func mkOracle(config config.OracleConfiguration, magic netmode.Magic, chain *core.Blockchain, serv *network.Server, log *zap.Logger) (*oracle.Oracle, error) {
if !config.Enabled {
return nil, nil
}
orcCfg := oracle.Config{
Log: log,
Network: magic,
MainCfg: config,
Chain: chain,
OnTransaction: serv.RelayTxn,
}
orc, err := oracle.NewOracle(orcCfg)
if err != nil {
return nil, fmt.Errorf("can't initialize Oracle module: %w", err)
}
chain.SetOracle(orc)
serv.AddService(orc)
return orc, nil
}
func mkConsensus(config config.Consensus, tpb time.Duration, chain *core.Blockchain, serv *network.Server, log *zap.Logger) (consensus.Service, error) {
if !config.Enabled {
return nil, nil
}
srv, err := consensus.NewService(consensus.Config{
Logger: log,
Broadcast: serv.BroadcastExtensible,
Chain: chain,
BlockQueue: serv.GetBlockQueue(),
ProtocolConfiguration: chain.GetConfig().ProtocolConfiguration,
RequestTx: serv.RequestTx,
StopTxFlow: serv.StopTxFlow,
Wallet: config.UnlockWallet,
TimePerBlock: tpb,
})
if err != nil {
return nil, fmt.Errorf("can't initialize Consensus module: %w", err)
}
serv.AddConsensusService(srv, srv.OnPayload, srv.OnTransaction)
return srv, nil
}
func mkP2PNotary(config config.P2PNotary, chain *core.Blockchain, serv *network.Server, log *zap.Logger) (*notary.Notary, error) {
if !config.Enabled {
return nil, nil
}
if !chain.P2PSigExtensionsEnabled() {
return nil, errors.New("P2PSigExtensions are disabled, but Notary service is enabled")
}
cfg := notary.Config{
MainCfg: config,
Chain: chain,
Log: log,
}
n, err := notary.NewNotary(cfg, serv.Net, serv.GetNotaryPool(), func(tx *transaction.Transaction) error {
err := serv.RelayTxn(tx)
if err != nil && !errors.Is(err, core.ErrAlreadyExists) {
return fmt.Errorf("can't relay completed notary transaction: hash %s, error: %w", tx.Hash().StringLE(), err)
}
return nil
})
if err != nil {
return nil, fmt.Errorf("failed to create Notary module: %w", err)
}
serv.AddService(n)
chain.SetNotary(n)
return n, nil
}
func startServer(ctx *cli.Context) error {
if err := cmdargs.EnsureNone(ctx); err != nil {
return err
}
cfg, err := options.GetConfigFromContext(ctx)
if err != nil {
return cli.NewExitError(err, 1)
}
var logDebug = ctx.Bool("debug")
log, logLevel, logCloser, err := options.HandleLoggingParams(logDebug, cfg.ApplicationConfiguration)
if err != nil {
return cli.NewExitError(err, 1)
}
if logCloser != nil {
defer func() { _ = logCloser() }()
}
grace, cancel := context.WithCancel(newGraceContext())
defer cancel()
serverConfig, err := network.NewServerConfig(cfg)
if err != nil {
return cli.NewExitError(err, 1)
}
chain, prometheus, pprof, err := initBCWithMetrics(cfg, log)
if err != nil {
return cli.NewExitError(err, 1)
}
defer func() {
pprof.ShutDown()
prometheus.ShutDown()
chain.Close()
}()
serv, err := network.NewServer(serverConfig, chain, chain.GetStateSyncModule(), log)
if err != nil {
return cli.NewExitError(fmt.Errorf("failed to create network server: %w", err), 1)
}
srMod := chain.GetStateModule().(*corestate.Module) // Take full responsibility here.
sr, err := stateroot.New(serverConfig.StateRootCfg, srMod, log, chain, serv.BroadcastExtensible)
if err != nil {
return cli.NewExitError(fmt.Errorf("can't initialize StateRoot service: %w", err), 1)
}
serv.AddExtensibleService(sr, stateroot.Category, sr.OnPayload)
oracleSrv, err := mkOracle(cfg.ApplicationConfiguration.Oracle, cfg.ProtocolConfiguration.Magic, chain, serv, log)
if err != nil {
return cli.NewExitError(err, 1)
}
dbftSrv, err := mkConsensus(cfg.ApplicationConfiguration.Consensus, serverConfig.TimePerBlock, chain, serv, log)
if err != nil {
return cli.NewExitError(err, 1)
}
p2pNotary, err := mkP2PNotary(cfg.ApplicationConfiguration.P2PNotary, chain, serv, log)
if err != nil {
return cli.NewExitError(err, 1)
}
errChan := make(chan error)
rpcServer := rpcsrv.New(chain, cfg.ApplicationConfiguration.RPC, serv, oracleSrv, log, errChan)
serv.AddService(&rpcServer)
go serv.Start(errChan)
if !cfg.ApplicationConfiguration.RPC.StartWhenSynchronized {
// Run RPC server in a separate routine. This is necessary to avoid a potential
// deadlock: Start() can write errors to errChan which is not yet read in the
// current execution context (see for-loop below).
go rpcServer.Start()
}
sigCh := make(chan os.Signal, 1)
signal.Notify(sigCh, sighup)
signal.Notify(sigCh, sigusr1)
signal.Notify(sigCh, sigusr2)
fmt.Fprintln(ctx.App.Writer, Logo())
fmt.Fprintln(ctx.App.Writer, serv.UserAgent)
fmt.Fprintln(ctx.App.Writer)
var shutdownErr error
Main:
for {
select {
case err := <-errChan:
shutdownErr = fmt.Errorf("server error: %w", err)
cancel()
case sig := <-sigCh:
var newLogLevel = zapcore.InvalidLevel
log.Info("signal received", zap.Stringer("name", sig))
cfgnew, err := options.GetConfigFromContext(ctx)
if err != nil {
log.Warn("can't reread the config file, signal ignored", zap.Error(err))
break // Continue working.
}
if !cfg.ProtocolConfiguration.Equals(&cfgnew.ProtocolConfiguration) {
log.Warn("ProtocolConfiguration changed, signal ignored")
break // Continue working.
}
if !cfg.ApplicationConfiguration.EqualsButServices(&cfgnew.ApplicationConfiguration) {
log.Warn("ApplicationConfiguration changed in incompatible way, signal ignored")
break // Continue working.
}
if !logDebug && cfgnew.ApplicationConfiguration.LogLevel != cfg.ApplicationConfiguration.LogLevel {
newLogLevel, err = zapcore.ParseLevel(cfgnew.ApplicationConfiguration.LogLevel)
if err != nil {
log.Warn("wrong LogLevel in ApplicationConfiguration, signal ignored", zap.Error(err))
break // Continue working.
}
}
configureAddresses(&cfgnew.ApplicationConfiguration)
switch sig {
case sighup:
if newLogLevel != zapcore.InvalidLevel {
logLevel.SetLevel(newLogLevel)
log.Warn("using new logging level", zap.Stringer("level", newLogLevel))
}
serv.DelService(&rpcServer)
rpcServer.Shutdown()
rpcServer = rpcsrv.New(chain, cfgnew.ApplicationConfiguration.RPC, serv, oracleSrv, log, errChan)
serv.AddService(&rpcServer)
if !cfgnew.ApplicationConfiguration.RPC.StartWhenSynchronized || serv.IsInSync() {
// Here similar to the initial run (see above for-loop), so async.
go rpcServer.Start()
}
pprof.ShutDown()
pprof = metrics.NewPprofService(cfgnew.ApplicationConfiguration.Pprof, log)
err = pprof.Start()
if err != nil {
shutdownErr = fmt.Errorf("failed to start Pprof service: %w", err)
cancel() // Fatal error, like for RPC server.
}
prometheus.ShutDown()
prometheus = metrics.NewPrometheusService(cfgnew.ApplicationConfiguration.Prometheus, log)
err = prometheus.Start()
if err != nil {
shutdownErr = fmt.Errorf("failed to start Prometheus service: %w", err)
cancel() // Fatal error, like for RPC server.
}
case sigusr1:
if oracleSrv != nil {
serv.DelService(oracleSrv)
chain.SetOracle(nil)
rpcServer.SetOracleHandler(nil)
oracleSrv.Shutdown()
}
oracleSrv, err = mkOracle(cfgnew.ApplicationConfiguration.Oracle, cfgnew.ProtocolConfiguration.Magic, chain, serv, log)
if err != nil {
log.Error("failed to create oracle service", zap.Error(err))
break // Keep going.
}
if oracleSrv != nil {
rpcServer.SetOracleHandler(oracleSrv)
if serv.IsInSync() {
oracleSrv.Start()
}
}
if p2pNotary != nil {
serv.DelService(p2pNotary)
chain.SetNotary(nil)
p2pNotary.Shutdown()
}
p2pNotary, err = mkP2PNotary(cfgnew.ApplicationConfiguration.P2PNotary, chain, serv, log)
if err != nil {
log.Error("failed to create notary service", zap.Error(err))
break // Keep going.
}
if p2pNotary != nil && serv.IsInSync() {
p2pNotary.Start()
}
serv.DelExtensibleService(sr, stateroot.Category)
srMod.SetUpdateValidatorsCallback(nil)
sr.Shutdown()
sr, err = stateroot.New(cfgnew.ApplicationConfiguration.StateRoot, srMod, log, chain, serv.BroadcastExtensible)
if err != nil {
log.Error("failed to create state validation service", zap.Error(err))
break // The show must go on.
}
serv.AddExtensibleService(sr, stateroot.Category, sr.OnPayload)
if serv.IsInSync() {
sr.Start()
}
case sigusr2:
if dbftSrv != nil {
serv.DelConsensusService(dbftSrv)
dbftSrv.Shutdown()
}
dbftSrv, err = mkConsensus(cfgnew.ApplicationConfiguration.Consensus, serverConfig.TimePerBlock, chain, serv, log)
if err != nil {
log.Error("failed to create consensus service", zap.Error(err))
break // Whatever happens, I'll leave it all to chance.
}
if dbftSrv != nil && serv.IsInSync() {
dbftSrv.Start()
}
}
cfg = cfgnew
case <-grace.Done():
signal.Stop(sigCh)
serv.Shutdown()
break Main
}
}
if shutdownErr != nil {
return cli.NewExitError(shutdownErr, 1)
}
return nil
}
// configureAddresses sets up addresses for RPC, Prometheus and Pprof depending from the provided config.
// In case RPC or Prometheus or Pprof Address provided each of them will use it.
// In case global Address (of the node) provided and RPC/Prometheus/Pprof don't have configured addresses they will
// use global one. So Node and RPC and Prometheus and Pprof will run on one address.
func configureAddresses(cfg *config.ApplicationConfiguration) {
if cfg.Address != nil && *cfg.Address != "" { //nolint:staticcheck // SA1019: cfg.Address is deprecated
if cfg.RPC.Address == nil || *cfg.RPC.Address == "" { //nolint:staticcheck // SA1019: cfg.RPC.Address is deprecated
cfg.RPC.Address = cfg.Address //nolint:staticcheck // SA1019: cfg.RPC.Address is deprecated
}
if cfg.Prometheus.Address == nil || *cfg.Prometheus.Address == "" { //nolint:staticcheck // SA1019: cfg.Prometheus.Address is deprecated
cfg.Prometheus.Address = cfg.Address //nolint:staticcheck // SA1019: cfg.Prometheus.Address is deprecated
}
if cfg.Pprof.Address == nil || *cfg.Pprof.Address == "" { //nolint:staticcheck // SA1019: cfg.Pprof.Address is deprecated
cfg.Pprof.Address = cfg.Address //nolint:staticcheck // SA1019: cfg.Pprof.Address is deprecated
}
}
}
// initBlockChain initializes BlockChain with preselected DB.
func initBlockChain(cfg config.Config, log *zap.Logger) (*core.Blockchain, storage.Store, error) {
store, err := storage.NewStore(cfg.ApplicationConfiguration.DBConfiguration)
if err != nil {
return nil, nil, cli.NewExitError(fmt.Errorf("could not initialize storage: %w", err), 1)
}
chain, err := core.NewBlockchain(store, cfg.Blockchain(), log)
if err != nil {
errText := "could not initialize blockchain: %w"
errArgs := []any{err}
closeErr := store.Close()
if closeErr != nil {
errText += "; failed to close the DB: %w"
errArgs = append(errArgs, closeErr)
}
return nil, nil, cli.NewExitError(fmt.Errorf(errText, errArgs...), 1)
}
return chain, store, nil
}
// Logo returns NeoGo logo.
func Logo() string {
return `
_ ____________ __________
/ | / / ____/ __ \ / ____/ __ \
/ |/ / __/ / / / /_____/ / __/ / / /
/ /| / /___/ /_/ /_____/ /_/ / /_/ /
/_/ |_/_____/\____/ \____/\____/
`
}