storj/pkg/kademlia/kademlia.go

// Copyright (C) 2019 Storj Labs, Inc.
// See LICENSE for copying information.

package kademlia

import (
	"context"
	"math/rand"
	"sync"
	"sync/atomic"
	"time"

	"github.com/zeebo/errs"
	"go.uber.org/zap"
	monkit "gopkg.in/spacemonkeygo/monkit.v2"

	"storj.io/storj/internal/sync2"
	"storj.io/storj/pkg/identity"
	"storj.io/storj/pkg/overlay"
	"storj.io/storj/pkg/pb"
	"storj.io/storj/pkg/storj"
	"storj.io/storj/pkg/transport"
	"storj.io/storj/storage"
)

var (
	// NodeErr is the class for all errors pertaining to node operations
	NodeErr = errs.Class("node error")
	// BootstrapErr is the class for all errors pertaining to bootstrapping a node
	BootstrapErr = errs.Class("bootstrap node error")
	// NodeNotFound is returned when a lookup can not produce the requested node
	NodeNotFound = errs.Class("node not found")
	// TODO: shouldn't default to TCP but not sure what to do yet
	defaultTransport = pb.NodeTransport_TCP_TLS_GRPC
	mon              = monkit.Package()
)

// Kademlia is an implementation of kademlia adhering to the DHT interface.
type Kademlia struct {
	log            *zap.Logger
	alpha          int // alpha is a system wide concurrency parameter
	routingTable   *RoutingTable
	bootstrapNodes []pb.Node
	dialer         *Dialer
	lookups        sync2.WorkGroup

	bootstrapFinished    sync2.Fence
	bootstrapBackoffMax  time.Duration
	bootstrapBackoffBase time.Duration

	refreshThreshold int64
	RefreshBuckets   sync2.Cycle

	mu          sync.Mutex
	lastPinged  time.Time
	lastQueried time.Time
}

// NewService returns a newly configured Kademlia instance
func NewService(log *zap.Logger, transport transport.Client, rt *RoutingTable, config Config) (*Kademlia, error) {
	k := &Kademlia{
		log:                  log,
		alpha:                config.Alpha,
		routingTable:         rt,
		bootstrapNodes:       config.BootstrapNodes(),
		bootstrapBackoffMax:  config.BootstrapBackoffMax,
		bootstrapBackoffBase: config.BootstrapBackoffBase,
		dialer:               NewDialer(log.Named("dialer"), transport),
		refreshThreshold:     int64(time.Minute),
	}

	return k, nil
}

// Close closes all kademlia connections and prevents new ones from being created.
func (k *Kademlia) Close() error {
	dialerErr := k.dialer.Close()
	k.lookups.Close()
	k.lookups.Wait()
	return dialerErr
}

// LastPinged returns last time someone pinged this node.
func (k *Kademlia) LastPinged() time.Time {
	k.mu.Lock()
	defer k.mu.Unlock()
	return k.lastPinged
}

// Pinged notifies the service it has been remotely pinged.
func (k *Kademlia) Pinged() {
	k.mu.Lock()
	defer k.mu.Unlock()
	k.lastPinged = time.Now()
}

// LastQueried returns last time someone queried this node.
func (k *Kademlia) LastQueried() time.Time {
	k.mu.Lock()
	defer k.mu.Unlock()
	return k.lastQueried
}

// Queried notifies the service it has been remotely queried
func (k *Kademlia) Queried() {
	k.mu.Lock()
	defer k.mu.Unlock()
	k.lastQueried = time.Now()
}

// FindNear returns all nodes from a starting node up to a maximum limit
// stored in the local routing table.
func (k *Kademlia) FindNear(ctx context.Context, start storj.NodeID, limit int) (_ []*pb.Node, err error) {
	defer mon.Task()(&ctx)(&err)
	return k.routingTable.FindNear(ctx, start, limit)
}

// GetBucketIds returns a storage.Keys type of bucket ID's in the Kademlia instance
func (k *Kademlia) GetBucketIds(ctx context.Context) (_ storage.Keys, err error) {
	defer mon.Task()(&ctx)(&err)
	return k.routingTable.GetBucketIds(ctx)
}

// Local returns the local node
func (k *Kademlia) Local() overlay.NodeDossier {
	return k.routingTable.Local()
}

// SetBootstrapNodes sets the bootstrap nodes.
// Must be called before anything starting to use kademlia.
func (k *Kademlia) SetBootstrapNodes(nodes []pb.Node) { k.bootstrapNodes = nodes }

// GetBootstrapNodes gets the bootstrap nodes.
func (k *Kademlia) GetBootstrapNodes() []pb.Node { return k.bootstrapNodes }

// DumpNodes returns all the nodes in the node database
func (k *Kademlia) DumpNodes(ctx context.Context) (_ []*pb.Node, err error) {
	defer mon.Task()(&ctx)(&err)
	return k.routingTable.DumpNodes(ctx)
}

// Bootstrap contacts one of a set of pre defined trusted nodes on the network and
// begins populating the local Kademlia node
func (k *Kademlia) Bootstrap(ctx context.Context) (err error) {
	defer mon.Task()(&ctx)(&err)
	defer k.bootstrapFinished.Release()

	if !k.lookups.Start() {
		return context.Canceled
	}
	defer k.lookups.Done()

	if len(k.bootstrapNodes) == 0 {
		k.log.Warn("No bootstrap address specified.")
		return nil
	}

	waitInterval := k.bootstrapBackoffBase

	var errGroup errs.Group
	for i := 0; waitInterval < k.bootstrapBackoffMax; i++ {
		if i > 0 {
			time.Sleep(waitInterval)
			waitInterval *= 2
		}

		var foundOnlineBootstrap bool
		for i, node := range k.bootstrapNodes {
			if ctx.Err() != nil {
				errGroup.Add(ctx.Err())
				return errGroup.Err()
			}

			ident, err := k.dialer.FetchPeerIdentityUnverified(ctx, node.Address.Address)
			if err != nil {
				errGroup.Add(err)
				continue
			}

			// FetchPeerIdentityUnverified uses transport.DialAddress, which should be
			// enough to have the TransportObservers find out about this node. Unfortunately,
			// getting DialAddress to be able to grab the node id seems challenging with gRPC.
			// The way FetchPeerIdentityUnverified does is is to do a basic ping request, which
			// we have now done. Let's tell all the transport observers now.
			// TODO: remove the explicit transport observer notification
			k.dialer.transport.AlertSuccess(ctx, &pb.Node{
				Id:      ident.ID,
				Address: node.Address,
			})

			k.routingTable.mutex.Lock()
			node.Id = ident.ID
			k.bootstrapNodes[i] = node
			k.routingTable.mutex.Unlock()
			foundOnlineBootstrap = true
		}

		if !foundOnlineBootstrap {
			errGroup.Add(Error.New("no bootstrap node found online"))
			continue
		}

		//find nodes most similar to self
		k.routingTable.mutex.Lock()
		id := k.routingTable.self.Id
		k.routingTable.mutex.Unlock()
		_, err := k.lookup(ctx, id)
		if err != nil {
			errGroup.Add(err)
			continue
		}
		return nil
		// TODO(dylan): We do not currently handle this last bit of behavior.
		// ```
		// Finally, u refreshes all k-buckets further away than its closest neighbor.
		// During the refreshes, u both populates its own k-buckets and inserts
		// itself into other nodes' k-buckets as necessary.
		// ```
	}

	errGroup.Add(Error.New("unable to start bootstrap after final wait time of %s", waitInterval))
	return errGroup.Err()
}

// WaitForBootstrap waits for bootstrap pinging has been completed.
func (k *Kademlia) WaitForBootstrap() {
	k.bootstrapFinished.Wait()
}

// FetchPeerIdentity connects to a node and returns its peer identity
func (k *Kademlia) FetchPeerIdentity(ctx context.Context, nodeID storj.NodeID) (_ *identity.PeerIdentity, err error) {
	defer mon.Task()(&ctx)(&err)
	if !k.lookups.Start() {
		return nil, context.Canceled
	}
	defer k.lookups.Done()
	node, err := k.FindNode(ctx, nodeID)
	if err != nil {
		return nil, err
	}
	return k.dialer.FetchPeerIdentity(ctx, node)
}

// Ping checks that the provided node is still accessible on the network
func (k *Kademlia) Ping(ctx context.Context, node pb.Node) (_ pb.Node, err error) {
	defer mon.Task()(&ctx)(&err)
	if !k.lookups.Start() {
		return pb.Node{}, context.Canceled
	}
	defer k.lookups.Done()

	ok, err := k.dialer.PingNode(ctx, node)
	if err != nil {
		return pb.Node{}, NodeErr.Wrap(err)
	}
	if !ok {
		return pb.Node{}, NodeErr.New("Failed pinging node")
	}
	return node, nil
}

// FetchInfo connects to a node address and returns the node info
func (k *Kademlia) FetchInfo(ctx context.Context, node pb.Node) (_ *pb.InfoResponse, err error) {
	defer mon.Task()(&ctx)(&err)
	if !k.lookups.Start() {
		return nil, context.Canceled
	}
	defer k.lookups.Done()

	info, err := k.dialer.FetchInfo(ctx, node)
	if err != nil {
		return nil, NodeErr.Wrap(err)
	}
	return info, nil
}

// FindNode looks up the provided NodeID first in the local Node, and if it is not found
// begins searching the network for the NodeID. Returns and error if node was not found
func (k *Kademlia) FindNode(ctx context.Context, nodeID storj.NodeID) (_ pb.Node, err error) {
	defer mon.Task()(&ctx)(&err)
	if !k.lookups.Start() {
		return pb.Node{}, context.Canceled
	}
	defer k.lookups.Done()

	results, err := k.lookup(ctx, nodeID)
	if err != nil {
		return pb.Node{}, err
	}
	if len(results) < 1 {
		return pb.Node{}, NodeNotFound.Wrap(err)
	}
	return *results[0], nil
}

//lookup initiates a kadmelia node lookup
func (k *Kademlia) lookup(ctx context.Context, nodeID storj.NodeID) (_ []*pb.Node, err error) {
	defer mon.Task()(&ctx)(&err)
	if !k.lookups.Start() {
		return nil, context.Canceled
	}
	defer k.lookups.Done()

	nodes, err := k.routingTable.FindNear(ctx, nodeID, k.routingTable.K())
	if err != nil {
		return nil, err
	}

	self := k.routingTable.Local().Node
	lookup := newPeerDiscovery(k.log, k.dialer, nodeID, nodes, k.routingTable.K(), k.alpha, &self)
	results, err := lookup.Run(ctx)
	if err != nil {
		return nil, err
	}
	bucket, err := k.routingTable.getKBucketID(ctx, nodeID)
	if err != nil {
		k.log.Warn("Error getting getKBucketID in kad lookup")
	} else {
		err = k.routingTable.SetBucketTimestamp(ctx, bucket[:], time.Now())
		if err != nil {
			k.log.Warn("Error updating bucket timestamp in kad lookup")
		}
	}
	return results, nil
}

// GetNodesWithinKBucket returns all the routing nodes in the specified k-bucket
func (k *Kademlia) GetNodesWithinKBucket(ctx context.Context, bID bucketID) (_ []*pb.Node, err error) {
	return k.routingTable.getUnmarshaledNodesFromBucket(ctx, bID)
}

// GetCachedNodesWithinKBucket returns all the cached nodes in the specified k-bucket
func (k *Kademlia) GetCachedNodesWithinKBucket(bID bucketID) []*pb.Node {
	return k.routingTable.replacementCache[bID]
}

// SetBucketRefreshThreshold changes the threshold when buckets are considered stale and need refreshing.
func (k *Kademlia) SetBucketRefreshThreshold(threshold time.Duration) {
	atomic.StoreInt64(&k.refreshThreshold, int64(threshold))
}

// Run occasionally refreshes stale kad buckets
func (k *Kademlia) Run(ctx context.Context) (err error) {
	defer mon.Task()(&ctx)(&err)

	if !k.lookups.Start() {
		return context.Canceled
	}
	defer k.lookups.Done()

	k.RefreshBuckets.SetInterval(5 * time.Minute)
	return k.RefreshBuckets.Run(ctx, func(ctx context.Context) error {
		threshold := time.Duration(atomic.LoadInt64(&k.refreshThreshold))
		err := k.refresh(ctx, threshold)
		if err != nil {
			k.log.Warn("bucket refresh failed", zap.Error(err))
		}
		return nil
	})
}

// refresh updates each Kademlia bucket not contacted in the last hour
func (k *Kademlia) refresh(ctx context.Context, threshold time.Duration) (err error) {
	defer mon.Task()(&ctx)(&err)
	bIDs, err := k.routingTable.GetBucketIds(ctx)
	if err != nil {
		return Error.Wrap(err)
	}
	now := time.Now()
	startID := bucketID{}
	var errors errs.Group
	for _, bID := range bIDs {
		endID := keyToBucketID(bID)
		ts, tErr := k.routingTable.GetBucketTimestamp(ctx, bID)
		if tErr != nil {
			errors.Add(tErr)
		} else if now.After(ts.Add(threshold)) {
			rID, _ := randomIDInRange(startID, endID)
			_, _ = k.FindNode(ctx, rID) // ignore node not found
		}
		startID = endID
	}
	return Error.Wrap(errors.Err())
}

// randomIDInRange finds a random node ID with a range (start..end]
func randomIDInRange(start, end bucketID) (storj.NodeID, error) {
	randID := storj.NodeID{}
	divergedHigh := false
	divergedLow := false
	for x := 0; x < len(randID); x++ {
		s := byte(0)
		if !divergedLow {
			s = start[x]
		}
		e := byte(255)
		if !divergedHigh {
			e = end[x]
		}
		if s > e {
			return storj.NodeID{}, errs.New("Random id range was invalid")
		}
		if s == e {
			randID[x] = s
		} else {
			r := s + byte(rand.Intn(int(e-s))) + 1
			if r < e {
				divergedHigh = true
			}
			if r > s {
				divergedLow = true
			}
			randID[x] = r
		}
	}
	if !divergedLow {
		if !divergedHigh { // start == end
			return storj.NodeID{}, errs.New("Random id range was invalid")
		} else if randID[len(randID)-1] == start[len(randID)-1] { // start == randID
			randID[len(randID)-1] = start[len(randID)-1] + 1
		}
	}
	return randID, nil
}