func newObcBatch(id uint64, config *viper.Viper, stack consensus.Stack) *obcBatch {
var err error
op := &obcBatch{
obcGeneric: obcGeneric{stack},
stack: stack,
}
op.persistForward.persistor = stack
logger.Debug("Replica %d obtaining startup information", id)
op.pbft = newPbftCore(id, config, op)
op.batchSize = config.GetInt("general.batchSize")
op.batchStore = nil
op.batchTimeout, err = time.ParseDuration(config.GetString("general.timeout.batch"))
if err != nil {
panic(fmt.Errorf("Cannot parse batch timeout: %s", err))
}
op.incomingChan = make(chan *batchMessage)
// create non-running timer
op.batchTimer = time.NewTimer(100 * time.Hour) // XXX ugly
op.batchTimer.Stop()
op.idleChan = make(chan struct{})
go op.main()
return op
}
func newObcBatch(id uint64, config *viper.Viper, stack consensus.Stack) *obcBatch {
var err error
op := &obcBatch{
obcGeneric: obcGeneric{stack: stack},
}
op.persistForward.persistor = stack
logger.Debugf("Replica %d obtaining startup information", id)
op.manager = events.NewManagerImpl() // TODO, this is hacky, eventually rip it out
op.manager.SetReceiver(op)
etf := events.NewTimerFactoryImpl(op.manager)
op.pbft = newPbftCore(id, config, op, etf)
op.manager.Start()
blockchainInfoBlob := stack.GetBlockchainInfoBlob()
op.externalEventReceiver.manager = op.manager
op.broadcaster = newBroadcaster(id, op.pbft.N, op.pbft.f, op.pbft.broadcastTimeout, stack)
op.manager.Queue() <- workEvent(func() {
op.pbft.stateTransfer(&stateUpdateTarget{
checkpointMessage: checkpointMessage{
seqNo: op.pbft.lastExec,
id: blockchainInfoBlob,
},
})
})
op.batchSize = config.GetInt("general.batchsize")
op.batchStore = nil
op.batchTimeout, err = time.ParseDuration(config.GetString("general.timeout.batch"))
if err != nil {
panic(fmt.Errorf("Cannot parse batch timeout: %s", err))
}
logger.Infof("PBFT Batch size = %d", op.batchSize)
logger.Infof("PBFT Batch timeout = %v", op.batchTimeout)
if op.batchTimeout >= op.pbft.requestTimeout {
op.pbft.requestTimeout = 3 * op.batchTimeout / 2
logger.Warningf("Configured request timeout must be greater than batch timeout, setting to %v", op.pbft.requestTimeout)
}
if op.pbft.requestTimeout >= op.pbft.nullRequestTimeout && op.pbft.nullRequestTimeout != 0 {
op.pbft.nullRequestTimeout = 3 * op.pbft.requestTimeout / 2
logger.Warningf("Configured null request timeout must be greater than request timeout, setting to %v", op.pbft.nullRequestTimeout)
}
op.incomingChan = make(chan *batchMessage)
op.batchTimer = etf.CreateTimer()
op.reqStore = newRequestStore()
op.deduplicator = newDeduplicator()
op.idleChan = make(chan struct{})
close(op.idleChan) // TODO remove eventually
return op
}
func newObcBatch(id uint64, config *viper.Viper, stack consensus.Stack) *obcBatch {
var err error
op := &obcBatch{stack: stack}
op.startup = make(chan []byte)
op.executor = NewOBCExecutor(config, op, stack)
logger.Debug("Replica %d obtaining startup information", id)
startupInfo := <-op.startup
close(op.startup)
op.pbft = newPbftCore(id, config, op, startupInfo)
queueSize := config.GetInt("executor.queuesize")
if queueSize <= int(op.pbft.L) {
logger.Error("Replica %d has executor queue size %d less than PBFT log size %d, this indicates a misconfiguration", id, queueSize, op.pbft.L)
}
op.batchSize = config.GetInt("general.batchSize")
op.batchStore = nil
op.batchTimeout, err = time.ParseDuration(config.GetString("general.timeout.batch"))
if err != nil {
panic(fmt.Errorf("Cannot parse batch timeout: %s", err))
}
// create non-running timer
op.batchTimer = time.NewTimer(100 * time.Hour) // XXX ugly
op.batchTimer.Stop()
go op.batchTimerHander()
return op
}
func NewOBCExecutor(config *viper.Viper, orderer Orderer, stack statetransfer.PartialStack) (obcex *obcExecutor) {
var err error
obcex = &obcExecutor{}
queueSize := config.GetInt("executor.queuesize")
if queueSize <= 0 {
panic("Queue size must be positive")
}
obcex.executionQueue = make(chan *transaction, queueSize)
obcex.syncTargets = make(chan *syncTarget)
obcex.completeSync = make(chan *syncTarget)
obcex.threadIdle = make(chan struct{})
obcex.threadExit = make(chan struct{})
obcex.orderer = orderer
obcex.executorStack = stack
obcex.id, _, err = stack.GetNetworkHandles()
if nil != err {
logger.Error("Could not resolve our own PeerID, assigning dummy")
obcex.id = &pb.PeerID{"Dummy"}
}
logger.Info("Executor for %v using queue size of %d", obcex.id, queueSize)
obcex.sts = statetransfer.NewStateTransferState(config, stack)
listener := struct{ statetransfer.ProtoListener }{}
listener.CompletedImpl = obcex.stateTransferCompleted
obcex.sts.RegisterListener(&listener)
go obcex.queueThread()
return
}
func loadOAuth2Config(v *viper.Viper) {
Config.OAuth2Enabled = v.GetBool("stormpath.web.oauth2.enabled")
Config.OAuth2URI = v.GetString("stormpath.web.oauth2.uri")
Config.OAuth2ClientCredentialsGrantTypeEnabled = v.GetBool("stormpath.web.oauth2.client_credentials.enabled")
Config.OAuth2ClientCredentialsGrantTypeAccessTokenTTL = time.Duration(v.GetInt("stormpath.web.oauth2.client_credentials.accessToken.ttl")) * time.Second
Config.OAuth2PasswordGrantTypeEnabled = v.GetBool("stormpath.web.oauth2.password.enabled")
Config.OAuth2PasswordGrantTypeValidationStrategy = v.GetString("stormpath.web.oauth2.password.validationStrategy")
}
func newInstance(v *viper.Viper) *instance {
var i = instance{
viper: v,
slots: v.GetInt("start.parallel"),
}
i.slotAvailable = sync.NewCond(&i.m)
i.taskFinished = sync.NewCond(&i.m)
return &i
}
func ThreadlessNewStateTransferState(config *viper.Viper, stack PartialStack) *StateTransferState {
var err error
sts := &StateTransferState{}
sts.stateTransferListenersLock = &sync.Mutex{}
sts.stack = stack
sts.id, _, err = stack.GetNetworkHandles()
if nil != err {
logger.Debug("Error resolving our own PeerID, this shouldn't happen")
sts.id = &protos.PeerID{"ERROR_RESOLVING_ID"}
}
sts.asynchronousTransferInProgress = false
sts.RecoverDamage = config.GetBool("statetransfer.recoverdamage")
sts.stateValid = true // Assume our starting state is correct unless told otherwise
sts.validBlockRanges = make([]*blockRange, 0)
sts.blockVerifyChunkSize = uint64(config.GetInt("statetransfer.blocksperrequest"))
if sts.blockVerifyChunkSize == 0 {
panic(fmt.Errorf("Must set statetransfer.blocksperrequest to be nonzero"))
}
sts.initiateStateSync = make(chan *syncMark)
sts.blockHashReceiver = make(chan *blockHashReply, 1)
sts.blockSyncReq = make(chan *blockSyncReq)
sts.threadExit = make(chan struct{})
sts.blockThreadIdle = true
sts.stateThreadIdle = true
sts.blockThreadIdleChan = make(chan struct{})
sts.stateThreadIdleChan = make(chan struct{})
sts.DiscoveryThrottleTime = 1 * time.Second // TODO make this configurable
sts.BlockRequestTimeout, err = time.ParseDuration(config.GetString("statetransfer.timeout.singleblock"))
if err != nil {
panic(fmt.Errorf("Cannot parse statetransfer.timeout.singleblock timeout: %s", err))
}
sts.StateDeltaRequestTimeout, err = time.ParseDuration(config.GetString("statetransfer.timeout.singlestatedelta"))
if err != nil {
panic(fmt.Errorf("Cannot parse statetransfer.timeout.singlestatedelta timeout: %s", err))
}
sts.StateSnapshotRequestTimeout, err = time.ParseDuration(config.GetString("statetransfer.timeout.fullstate"))
if err != nil {
panic(fmt.Errorf("Cannot parse statetransfer.timeout.fullstate timeout: %s", err))
}
return sts
}
func newPbftCore(id uint64, config *viper.Viper, consumer innerCPI) *pbftCore {
instance := &pbftCore{}
instance.id = id
instance.consumer = consumer
// in dev/debugging mode you are expected to override the config values
// with the environment variable OPENCHAIN_OBCPBFT_X_Y
// read from the config file
// you can override the config values with the environment variable prefix
// OPENCHAIN_OBCPBFT, e.g. OPENCHAIN_OBCPBFT_BYZANTINE
var err error
instance.byzantine = config.GetBool("replica.byzantine")
instance.f = config.GetInt("general.f")
instance.K = uint64(config.GetInt("general.K"))
instance.requestTimeout, err = time.ParseDuration(config.GetString("general.timeout.request"))
if err != nil {
panic(fmt.Errorf("Cannot parse request timeout: %s", err))
}
instance.newViewTimeout, err = time.ParseDuration(config.GetString("general.timeout.viewchange"))
if err != nil {
panic(fmt.Errorf("Cannot parse new view timeout: %s", err))
}
instance.activeView = true
instance.L = 2 * instance.K // log size
instance.replicaCount = 3*instance.f + 1
// init the logs
instance.certStore = make(map[msgID]*msgCert)
instance.reqStore = make(map[string]*Request)
instance.checkpointStore = make(map[Checkpoint]bool)
instance.chkpts = make(map[uint64]string)
instance.viewChangeStore = make(map[vcidx]*ViewChange)
instance.pset = make(map[uint64]*ViewChange_PQ)
instance.qset = make(map[qidx]*ViewChange_PQ)
instance.newViewStore = make(map[uint64]*NewView)
// load genesis checkpoint
stateHash, err := instance.consumer.getStateHash(0)
if err != nil {
panic(fmt.Errorf("Cannot load genesis block: %s", err))
}
instance.chkpts[0] = base64.StdEncoding.EncodeToString(stateHash)
// create non-running timer XXX ugly
instance.newViewTimer = time.NewTimer(100 * time.Hour)
instance.newViewTimer.Stop()
instance.lastNewViewTimeout = instance.newViewTimeout
instance.outstandingReqs = make(map[string]*Request)
go instance.timerHander()
return instance
}
func newObcBatch(id uint64, config *viper.Viper, stack consensus.Stack) *obcBatch {
var err error
op := &obcBatch{stack: stack}
op.pbft = newPbftCore(id, config, op, stack)
op.batchSize = config.GetInt("general.batchSize")
op.batchStore = nil
op.batchTimeout, err = time.ParseDuration(config.GetString("general.timeout.batch"))
if err != nil {
panic(fmt.Errorf("Cannot parse batch timeout: %s", err))
}
// create non-running timer
op.batchTimer = time.NewTimer(100 * time.Hour) // XXX ugly
op.batchTimer.Stop()
go op.batchTimerHander()
return op
}
func newObcBatch(id uint64, config *viper.Viper, cpi consensus.CPI) *obcBatch {
var err error
op := &obcBatch{cpi: cpi}
op.pbft = newPbftCore(id, config, op, cpi)
op.batchSize = config.GetInt("general.batchSize")
op.batchStore = make(map[string]*Request)
op.batchTimeout, err = time.ParseDuration(config.GetString("general.timeout.batch"))
if err != nil {
panic(fmt.Errorf("Cannot parse batch timeout: %s", err))
}
// create non-running timer XXX ugly
op.batchTimer = time.NewTimer(100 * time.Hour)
op.batchTimer.Stop()
go op.batchTimerHander()
return op
}
func ThreadlessNewStateTransferState(id *protos.PeerID, config *viper.Viper, ledger consensus.LedgerStack, defaultPeerIDs []*protos.PeerID) *StateTransferState {
sts := &StateTransferState{}
sts.stateTransferListenersLock = &sync.Mutex{}
sts.ledger = ledger
sts.id = id
sts.asynchronousTransferInProgress = false
logger.Debug("%v assigning %v to defaultPeerIDs", id, defaultPeerIDs)
sts.defaultPeerIDs = defaultPeerIDs
sts.RecoverDamage = config.GetBool("statetransfer.recoverdamage")
sts.stateValid = true // Assume our starting state is correct unless told otherwise
sts.validBlockRanges = make([]*blockRange, 0)
sts.blockVerifyChunkSize = uint64(config.GetInt("statetransfer.blocksperrequest"))
if sts.blockVerifyChunkSize == 0 {
panic(fmt.Errorf("Must set statetransfer.blocksperrequest to be nonzero"))
}
sts.initiateStateSync = make(chan *syncMark, 1)
sts.blockHashReceiver = make(chan *blockHashReply, 1)
sts.blockSyncReq = make(chan *blockSyncReq)
sts.blockThreadExit = make(chan struct{}, 1)
sts.stateThreadExit = make(chan struct{}, 1)
var err error
sts.BlockRequestTimeout, err = time.ParseDuration(config.GetString("statetransfer.timeout.singleblock"))
if err != nil {
panic(fmt.Errorf("Cannot parse statetransfer.timeout.singleblock timeout: %s", err))
}
sts.StateDeltaRequestTimeout, err = time.ParseDuration(config.GetString("statetransfer.timeout.singlestatedelta"))
if err != nil {
panic(fmt.Errorf("Cannot parse statetransfer.timeout.singlestatedelta timeout: %s", err))
}
sts.StateSnapshotRequestTimeout, err = time.ParseDuration(config.GetString("statetransfer.timeout.fullstate"))
if err != nil {
panic(fmt.Errorf("Cannot parse statetransfer.timeout.fullstate timeout: %s", err))
}
return sts
}
func newObcClassic(id uint64, config *viper.Viper, stack consensus.Stack) *obcClassic {
op := &obcClassic{stack: stack}
op.startup = make(chan []byte)
op.executor = NewOBCExecutor(config, op, stack)
logger.Debug("Replica %d obtaining startup information", id)
startupInfo := <-op.startup
close(op.startup)
op.pbft = newPbftCore(id, config, op, startupInfo)
queueSize := config.GetInt("executor.queuesize")
if queueSize <= int(op.pbft.L) {
logger.Error("Replica %d has executor queue size %d less than PBFT log size %d, this indicates a misconfiguration", id, queueSize, op.pbft.L)
}
return op
}
func newObcBatch(id uint64, config *viper.Viper, stack consensus.Stack) *obcBatch {
var err error
op := &obcBatch{
obcGeneric: obcGeneric{stack: stack},
}
op.persistForward.persistor = stack
logger.Debugf("Replica %d obtaining startup information", id)
op.manager = events.NewManagerImpl() // TODO, this is hacky, eventually rip it out
op.manager.SetReceiver(op)
etf := events.NewTimerFactoryImpl(op.manager)
op.pbft = newPbftCore(id, config, op, etf)
op.manager.Start()
op.externalEventReceiver.manager = op.manager
op.broadcaster = newBroadcaster(id, op.pbft.N, op.pbft.f, stack)
op.batchSize = config.GetInt("general.batchsize")
op.batchStore = nil
op.batchTimeout, err = time.ParseDuration(config.GetString("general.timeout.batch"))
if err != nil {
panic(fmt.Errorf("Cannot parse batch timeout: %s", err))
}
logger.Infof("PBFT Batch size = %d", op.batchSize)
logger.Infof("PBFT Batch timeout = %v", op.batchTimeout)
op.incomingChan = make(chan *batchMessage)
op.batchTimer = etf.CreateTimer()
op.outstandingReqs = make(map[*Request]struct{})
op.idleChan = make(chan struct{})
close(op.idleChan) // TODO remove eventually
return op
}
func getGlobalProject(v *viper.Viper) (*libcentrifugo.Project, bool) {
p := &libcentrifugo.Project{}
// TODO: the same as for structureFromConfig function
if v == nil {
if !viper.IsSet("project_name") || viper.GetString("project_name") == "" {
return nil, false
}
p.Name = libcentrifugo.ProjectKey(viper.GetString("project_name"))
p.Secret = viper.GetString("project_secret")
p.ConnLifetime = int64(viper.GetInt("project_connection_lifetime"))
p.Anonymous = viper.GetBool("project_anonymous")
p.Watch = viper.GetBool("project_watch")
p.Publish = viper.GetBool("project_publish")
p.JoinLeave = viper.GetBool("project_join_leave")
p.Presence = viper.GetBool("project_presence")
p.HistorySize = int64(viper.GetInt("project_history_size"))
p.HistoryLifetime = int64(viper.GetInt("project_history_lifetime"))
} else {
if !v.IsSet("project_name") || v.GetString("project_name") == "" {
return nil, false
}
p.Name = libcentrifugo.ProjectKey(v.GetString("project_name"))
p.Secret = v.GetString("project_secret")
p.ConnLifetime = int64(v.GetInt("project_connection_lifetime"))
p.Anonymous = v.GetBool("project_anonymous")
p.Watch = v.GetBool("project_watch")
p.Publish = v.GetBool("project_publish")
p.JoinLeave = v.GetBool("project_join_leave")
p.Presence = v.GetBool("project_presence")
p.HistorySize = int64(v.GetInt("project_history_size"))
p.HistoryLifetime = int64(v.GetInt("project_history_lifetime"))
}
var nl []libcentrifugo.Namespace
if v == nil {
viper.MarshalKey("project_namespaces", &nl)
} else {
v.MarshalKey("project_namespaces", &nl)
}
p.Namespaces = nl
return p, true
}
func newPbftCore(id uint64, config *viper.Viper, consumer innerStack) *pbftCore {
var err error
instance := &pbftCore{}
instance.id = id
instance.consumer = consumer
instance.closed = make(chan struct{})
instance.incomingChan = make(chan *pbftMessage)
instance.stateUpdatedChan = make(chan *checkpointMessage)
instance.stateUpdatingChan = make(chan *checkpointMessage)
instance.execCompleteChan = make(chan struct{})
instance.idleChan = make(chan struct{})
instance.injectChan = make(chan func())
// TODO Ultimately, the timer factory will be passed in, and the existence of the manager
// will be hidden from pbftCore, but in the interest of a small PR, leaving it here for now
instance.manager = newEventManagerImpl(instance)
etf := newEventTimerFactoryImpl(instance.manager)
instance.newViewTimer = etf.createTimer()
instance.N = config.GetInt("general.N")
instance.f = config.GetInt("general.f")
if instance.f*3+1 > instance.N {
panic(fmt.Sprintf("need at least %d enough replicas to tolerate %d byzantine faults, but only %d replicas configured", instance.f*3+1, instance.f, instance.N))
}
instance.K = uint64(config.GetInt("general.K"))
instance.logMultiplier = uint64(config.GetInt("general.logmultiplier"))
if instance.logMultiplier < 2 {
panic("Log multiplier must be greater than or equal to 2")
}
instance.L = instance.logMultiplier * instance.K // log size
instance.byzantine = config.GetBool("general.byzantine")
instance.requestTimeout, err = time.ParseDuration(config.GetString("general.timeout.request"))
if err != nil {
panic(fmt.Errorf("Cannot parse request timeout: %s", err))
}
instance.newViewTimeout, err = time.ParseDuration(config.GetString("general.timeout.viewchange"))
if err != nil {
panic(fmt.Errorf("Cannot parse new view timeout: %s", err))
}
instance.activeView = true
instance.replicaCount = instance.N
logger.Info("PBFT type = %T", instance.consumer)
logger.Info("PBFT Max number of validating peers (N) = %v", instance.N)
logger.Info("PBFT Max number of failing peers (f) = %v", instance.f)
logger.Info("PBFT byzantine flag = %v", instance.byzantine)
logger.Info("PBFT request timeout = %v", instance.requestTimeout)
logger.Info("PBFT view change timeout = %v", instance.newViewTimeout)
logger.Info("PBFT Checkpoint period (K) = %v", instance.K)
logger.Info("PBFT Log multiplier = %v", instance.logMultiplier)
logger.Info("PBFT log size (L) = %v", instance.L)
// init the logs
instance.certStore = make(map[msgID]*msgCert)
instance.reqStore = make(map[string]*Request)
instance.checkpointStore = make(map[Checkpoint]bool)
instance.chkpts = make(map[uint64]string)
instance.viewChangeStore = make(map[vcidx]*ViewChange)
instance.pset = make(map[uint64]*ViewChange_PQ)
instance.qset = make(map[qidx]*ViewChange_PQ)
instance.newViewStore = make(map[uint64]*NewView)
// initialize state transfer
instance.hChkpts = make(map[uint64]uint64)
instance.chkpts[0] = "XXX GENESIS"
instance.lastNewViewTimeout = instance.newViewTimeout
instance.outstandingReqs = make(map[string]*Request)
instance.missingReqs = make(map[string]bool)
instance.restoreState()
return instance
}
func newPbftCore(id uint64, config *viper.Viper, consumer innerStack, ledger consensus.LedgerStack) *pbftCore {
var err error
instance := &pbftCore{}
instance.id = id
instance.consumer = consumer
instance.ledger = ledger
instance.closed = make(chan bool)
instance.notifyCommit = make(chan bool, 1)
instance.notifyExec = sync.NewCond(&instance.internalLock)
instance.N = config.GetInt("general.N")
instance.f = config.GetInt("general.f")
if instance.f*3+1 > instance.N {
panic(fmt.Sprintf("need at least %d enough replicas to tolerate %d byzantine faults, but only %d replicas configured", instance.f*3+1, instance.f, instance.N))
}
instance.K = uint64(config.GetInt("general.K"))
instance.byzantine = config.GetBool("general.byzantine")
instance.requestTimeout, err = time.ParseDuration(config.GetString("general.timeout.request"))
if err != nil {
panic(fmt.Errorf("Cannot parse request timeout: %s", err))
}
instance.newViewTimeout, err = time.ParseDuration(config.GetString("general.timeout.viewchange"))
if err != nil {
panic(fmt.Errorf("Cannot parse new view timeout: %s", err))
}
instance.activeView = true
instance.L = 2 * instance.K // log size
instance.replicaCount = instance.N
// init the logs
instance.certStore = make(map[msgID]*msgCert)
instance.reqStore = make(map[string]*Request)
instance.checkpointStore = make(map[Checkpoint]bool)
instance.chkpts = make(map[uint64]*blockState)
instance.viewChangeStore = make(map[vcidx]*ViewChange)
instance.pset = make(map[uint64]*ViewChange_PQ)
instance.qset = make(map[qidx]*ViewChange_PQ)
instance.newViewStore = make(map[uint64]*NewView)
// initialize state transfer
instance.hChkpts = make(map[uint64]uint64)
defaultPeerIDs := make([]*protos.PeerID, instance.replicaCount-1)
if instance.replicaCount > 1 {
// For some tests, only 1 replica will be present, and defaultPeerIDs makes no sense
for i := uint64(0); i < uint64(instance.replicaCount); i++ {
handle, err := getValidatorHandle(i)
if err != nil {
panic(fmt.Errorf("Cannot retrieve handle for peer which must exist : %s", err))
}
if i < instance.id {
logger.Debug("Replica %d assigning %v to index %d for replicaCount %d and id %d", instance.id, handle, i, instance.replicaCount, instance.id)
defaultPeerIDs[i] = handle
} else if i > instance.id {
logger.Debug("Replica %d assigning %v to index %d for replicaCount %d and id %d", instance.id, handle, i-1, instance.replicaCount, instance.id)
defaultPeerIDs[i-1] = handle
} else {
// This is our ID, do not add it to the list of default peers
}
}
} else {
logger.Debug("Replica %d not initializing defaultPeerIDs, as replicaCount is %d", instance.id, instance.replicaCount)
}
if myHandle, err := getValidatorHandle(instance.id); err != nil {
panic("Could not retrieve own handle")
} else {
instance.sts = statetransfer.NewStateTransferState(myHandle, config, ledger, defaultPeerIDs)
}
listener := struct{ statetransfer.ProtoListener }{}
listener.CompletedImpl = instance.stateTransferCompleted
instance.sts.RegisterListener(&listener)
// load genesis checkpoint
genesisBlock, err := instance.ledger.GetBlock(0)
if err != nil {
panic(fmt.Errorf("Cannot load genesis block: %s", err))
}
genesisHash, err := ledger.HashBlock(genesisBlock)
if err != nil {
panic(fmt.Errorf("Cannot hash genesis block: %s", err))
}
instance.chkpts[0] = &blockState{
blockNumber: 0,
blockHash: base64.StdEncoding.EncodeToString(genesisHash),
}
// create non-running timer XXX ugly
instance.newViewTimer = time.NewTimer(100 * time.Hour)
instance.newViewTimer.Stop()
instance.timerResetCount = 1
instance.lastNewViewTimeout = instance.newViewTimeout
instance.outstandingReqs = make(map[string]*Request)
instance.missingReqs = make(map[string]bool)
go instance.timerHander()
go instance.executeRoutine()
return instance
}
func newPbftCore(id uint64, config *viper.Viper, consumer innerStack, etf events.TimerFactory) *pbftCore {
var err error
instance := &pbftCore{}
instance.id = id
instance.consumer = consumer
instance.newViewTimer = etf.CreateTimer()
instance.vcResendTimer = etf.CreateTimer()
instance.nullRequestTimer = etf.CreateTimer()
instance.N = config.GetInt("general.N")
instance.f = config.GetInt("general.f")
if instance.f*3+1 > instance.N {
panic(fmt.Sprintf("need at least %d enough replicas to tolerate %d byzantine faults, but only %d replicas configured", instance.f*3+1, instance.f, instance.N))
}
instance.K = uint64(config.GetInt("general.K"))
instance.logMultiplier = uint64(config.GetInt("general.logmultiplier"))
if instance.logMultiplier < 2 {
panic("Log multiplier must be greater than or equal to 2")
}
instance.L = instance.logMultiplier * instance.K // log size
instance.viewChangePeriod = uint64(config.GetInt("general.viewchangeperiod"))
instance.byzantine = config.GetBool("general.byzantine")
instance.requestTimeout, err = time.ParseDuration(config.GetString("general.timeout.request"))
if err != nil {
panic(fmt.Errorf("Cannot parse request timeout: %s", err))
}
instance.vcResendTimeout, err = time.ParseDuration(config.GetString("general.timeout.resendviewchange"))
if err != nil {
panic(fmt.Errorf("Cannot parse request timeout: %s", err))
}
instance.newViewTimeout, err = time.ParseDuration(config.GetString("general.timeout.viewchange"))
if err != nil {
panic(fmt.Errorf("Cannot parse new view timeout: %s", err))
}
instance.nullRequestTimeout, err = time.ParseDuration(config.GetString("general.timeout.nullrequest"))
if err != nil {
instance.nullRequestTimeout = 0
}
instance.broadcastTimeout, err = time.ParseDuration(config.GetString("general.timeout.broadcast"))
if err != nil {
panic(fmt.Errorf("Cannot parse new broadcast timeout: %s", err))
}
instance.activeView = true
instance.replicaCount = instance.N
logger.Infof("PBFT type = %T", instance.consumer)
logger.Infof("PBFT Max number of validating peers (N) = %v", instance.N)
logger.Infof("PBFT Max number of failing peers (f) = %v", instance.f)
logger.Infof("PBFT byzantine flag = %v", instance.byzantine)
logger.Infof("PBFT request timeout = %v", instance.requestTimeout)
logger.Infof("PBFT view change timeout = %v", instance.newViewTimeout)
logger.Infof("PBFT Checkpoint period (K) = %v", instance.K)
logger.Infof("PBFT broadcast timeout = %v", instance.broadcastTimeout)
logger.Infof("PBFT Log multiplier = %v", instance.logMultiplier)
logger.Infof("PBFT log size (L) = %v", instance.L)
if instance.nullRequestTimeout > 0 {
logger.Infof("PBFT null requests timeout = %v", instance.nullRequestTimeout)
} else {
logger.Infof("PBFT null requests disabled")
}
if instance.viewChangePeriod > 0 {
logger.Infof("PBFT view change period = %v", instance.viewChangePeriod)
} else {
logger.Infof("PBFT automatic view change disabled")
}
// init the logs
instance.certStore = make(map[msgID]*msgCert)
instance.reqBatchStore = make(map[string]*RequestBatch)
instance.checkpointStore = make(map[Checkpoint]bool)
instance.chkpts = make(map[uint64]string)
instance.viewChangeStore = make(map[vcidx]*ViewChange)
instance.pset = make(map[uint64]*ViewChange_PQ)
instance.qset = make(map[qidx]*ViewChange_PQ)
instance.newViewStore = make(map[uint64]*NewView)
// initialize state transfer
instance.hChkpts = make(map[uint64]uint64)
instance.chkpts[0] = "XXX GENESIS"
instance.lastNewViewTimeout = instance.newViewTimeout
instance.outstandingReqBatches = make(map[string]*RequestBatch)
instance.missingReqBatches = make(map[string]bool)
instance.restoreState()
instance.viewChangeSeqNo = ^uint64(0) // infinity
instance.updateViewChangeSeqNo()
return instance
}
func newPbftCore(id uint64, config *viper.Viper, consumer innerStack, startupID []byte) *pbftCore {
var err error
instance := &pbftCore{}
instance.id = id
instance.consumer = consumer
instance.closed = make(chan bool)
instance.notifyCommit = make(chan bool, 1)
instance.N = config.GetInt("general.N")
instance.f = config.GetInt("general.f")
if instance.f*3+1 > instance.N {
panic(fmt.Sprintf("need at least %d enough replicas to tolerate %d byzantine faults, but only %d replicas configured", instance.f*3+1, instance.f, instance.N))
}
instance.K = uint64(config.GetInt("general.K"))
instance.logMultiplier = uint64(config.GetInt("general.logmultiplier"))
instance.byzantine = config.GetBool("general.byzantine")
instance.requestTimeout, err = time.ParseDuration(config.GetString("general.timeout.request"))
if err != nil {
panic(fmt.Errorf("Cannot parse request timeout: %s", err))
}
instance.newViewTimeout, err = time.ParseDuration(config.GetString("general.timeout.viewchange"))
if err != nil {
panic(fmt.Errorf("Cannot parse new view timeout: %s", err))
}
instance.activeView = true
instance.L = instance.logMultiplier * instance.K // log size
instance.replicaCount = instance.N
logger.Info("PBFT type = %T", instance.consumer)
logger.Info("PBFT Max number of validating peers (N) = %v", instance.N)
logger.Info("PBFT Max number of failing peers (f) = %v", instance.f)
logger.Info("PBFT byzantine flag = %v", instance.byzantine)
logger.Info("PBFT request timeout = %v", instance.requestTimeout)
logger.Info("PBFT view change timeout = %v", instance.newViewTimeout)
logger.Info("PBFT Checkpoint period (K) = %v", instance.K)
logger.Info("PBFT Log multiplier = %v", instance.logMultiplier)
logger.Info("PBFT log size (L) = %v", instance.L)
// init the logs
instance.certStore = make(map[msgID]*msgCert)
instance.reqStore = make(map[string]*Request)
instance.checkpointStore = make(map[Checkpoint]bool)
instance.chkpts = make(map[uint64]string)
instance.viewChangeStore = make(map[vcidx]*ViewChange)
instance.pset = make(map[uint64]*ViewChange_PQ)
instance.qset = make(map[qidx]*ViewChange_PQ)
instance.newViewStore = make(map[uint64]*NewView)
// initialize state transfer
instance.hChkpts = make(map[uint64]uint64)
instance.chkpts[0] = base64.StdEncoding.EncodeToString(startupID)
// create non-running timer XXX ugly
instance.newViewTimer = time.NewTimer(100 * time.Hour)
instance.newViewTimer.Stop()
instance.timerResetCount = 1
instance.lastNewViewTimeout = instance.newViewTimeout
instance.outstandingReqs = make(map[string]*Request)
instance.missingReqs = make(map[string]bool)
go instance.timerHander()
return instance
}
func loadConfig(v *viper.Viper) *runnerConf {
v.SetDefault("check_timeout", 10)
v.SetDefault("host_window", 60)
v.SetDefault("host_threshold", 5)
v.SetDefault("flood_window", 120)
v.SetDefault("flood_threshold", 100)
v.SetDefault("flap_window", 1200)
v.SetDefault("flap_threshold", 5)
v.SetDefault("alert_threshold", 3)
v.SetDefault("worker_id", "worker1")
v.SetDefault("check_key", "canhazstatus")
conf := &runnerConf{}
conf.checkTimeout = v.GetDuration("check_timeout")
conf.hostWindow = int64(v.GetInt("host_window"))
conf.hostThreshold = v.GetInt("host_threshold")
conf.floodWindow = int64(v.GetInt("flood_window"))
conf.floodThreshold = v.GetInt("flood_threshold")
conf.flapWindow = v.GetInt("flap_window")
conf.flapThreshold = v.GetInt("flap_threshold")
conf.alertThreshold = v.GetInt("alert_threshold")
conf.workerQueue = v.GetString("worker_id")
conf.checkKey = v.GetString("check_key")
//twilio config
v.SetDefault("twilio_enable", false)
conf.twilioEnabled = v.GetBool("twilio_enable")
conf.twsid = v.GetString("twiliosid")
conf.twtoken = v.GetString("twiliotoken")
conf.twfrom = v.GetString("twiliofrom")
conf.twdest = v.GetStringSlice("twiliodest")
//pagerduty config
v.SetDefault("pagerduty_enable", false)
conf.pagerDutyEnabled = v.GetBool("pagerduty_enable")
conf.pagerDutyPriOneKey = v.GetString("pagerduty_priority_one_key")
conf.pagerDutyPriTwoKey = v.GetString("pagerduty_priority_two_key")
conf.pagerDutyIncidentKeyPrefix = v.GetString("pagerduty_incident_key_prefix")
return conf
}