lnd.xprv/routing/missioncontrol.go
Olaoluwa Osuntokun f494433cbf
routing+server: add new QueryBandwidth method to reduce outbound failures
In this commit, we introduce a new method to the channel router's config
struct: QueryBandwidth. This method allows the channel router to query
for the up-to-date available bandwidth of a particular link. In the case
that this link emanates from/to us, then we can query the switch to see
if the link is active (if not bandwidth is zero), and return the current
best estimate for the available bandwidth of the link. If the link,
isn't one of ours, then we can thread through the total maximal
capacity of the link.

In order to implement this, the missionControl struct will now query the
switch upon creation to obtain a fresh bandwidth snapshot. We take care
to do this in a distinct db transaction in order to now introduced a
circular waiting condition between the mutexes in bolt, and the channel
state machine.

The aim of this change is to reduce the number of unnecessary failures
during HTLC payment routing as we'll now skip any links that are
inactive, or just don't have enough bandwidth for the payment. Nodes
that have several hundred channels (all of which in various states of
activity and available bandwidth) should see a nice gain from this w.r.t
payment latency.
2018-05-14 16:23:54 -07:00

374 lines
14 KiB
Go

package routing
import (
"sync"
"time"
"github.com/coreos/bbolt"
"github.com/lightningnetwork/lnd/channeldb"
"github.com/lightningnetwork/lnd/lnwire"
"github.com/roasbeef/btcd/btcec"
)
const (
// vertexDecay is the decay period of colored vertexes added to
// missionControl. Once vertexDecay passes after an entry has been
// added to the prune view, it is garbage collected. This value is
// larger than edgeDecay as an edge failure typical indicates an
// unbalanced channel, while a vertex failure indicates a node is not
// online and active.
vertexDecay = time.Duration(time.Minute * 5)
// edgeDecay is the decay period of colored edges added to
// missionControl. Once edgeDecay passed after an entry has been added,
// it is garbage collected. This value is smaller than vertexDecay as
// an edge related failure during payment sending typically indicates
// that a channel was unbalanced, a condition which may quickly change.
//
// TODO(roasbeef): instead use random delay on each?
edgeDecay = time.Duration(time.Second * 5)
)
// missionControl contains state which summarizes the past attempts of HTLC
// routing by external callers when sending payments throughout the network.
// missionControl remembers the outcome of these past routing attempts (success
// and failure), and is able to provide hints/guidance to future HTLC routing
// attempts. missionControl maintains a decaying network view of the
// edges/vertexes that should be marked as "pruned" during path finding. This
// graph view acts as a shared memory during HTLC payment routing attempts.
// With each execution, if an error is encountered, based on the type of error
// and the location of the error within the route, an edge or vertex is added
// to the view. Later sending attempts will then query the view for all the
// vertexes/edges that should be ignored. Items in the view decay after a set
// period of time, allowing the view to be dynamic w.r.t network changes.
type missionControl struct {
// failedEdges maps a short channel ID to be pruned, to the time that
// it was added to the prune view. Edges are added to this map if a
// caller reports to missionControl a failure localized to that edge
// when sending a payment.
failedEdges map[uint64]time.Time
// failedVertexes maps a node's public key that should be pruned, to
// the time that it was added to the prune view. Vertexes are added to
// this map if a caller reports to missionControl a failure localized
// to that particular vertex.
failedVertexes map[Vertex]time.Time
graph *channeldb.ChannelGraph
selfNode *channeldb.LightningNode
queryBandwidth func(*channeldb.ChannelEdgeInfo) lnwire.MilliSatoshi
sync.Mutex
// TODO(roasbeef): further counters, if vertex continually unavailable,
// add to another generation
// TODO(roasbeef): also add favorable metrics for nodes
}
// newMissionControl returns a new instance of missionControl.
//
// TODO(roasbeef): persist memory
func newMissionControl(
g *channeldb.ChannelGraph, selfNode *channeldb.LightningNode,
qb func(*channeldb.ChannelEdgeInfo) lnwire.MilliSatoshi) *missionControl {
return &missionControl{
failedEdges: make(map[uint64]time.Time),
failedVertexes: make(map[Vertex]time.Time),
selfNode: selfNode,
queryBandwidth: qb,
graph: g,
}
}
// graphPruneView is a filter of sorts that path finding routines should
// consult during the execution. Any edges or vertexes within the view should
// be ignored during path finding. The contents of the view reflect the current
// state of the wider network from the PoV of mission control compiled via HTLC
// routing attempts in the past.
type graphPruneView struct {
edges map[uint64]struct{}
vertexes map[Vertex]struct{}
}
// GraphPruneView returns a new graphPruneView instance which is to be
// consulted during path finding. If a vertex/edge is found within the returned
// prune view, it is to be ignored as a goroutine has had issues routing
// through it successfully. Within this method the main view of the
// missionControl is garbage collected as entries are detected to be "stale".
func (m *missionControl) GraphPruneView() graphPruneView {
// First, we'll grab the current time, this value will be used to
// determine if an entry is stale or not.
now := time.Now()
m.Lock()
// For each of the vertexes that have been added to the prune view, if
// it is now "stale", then we'll ignore it and avoid adding it to the
// view we'll return.
vertexes := make(map[Vertex]struct{})
for vertex, pruneTime := range m.failedVertexes {
if now.Sub(pruneTime) >= vertexDecay {
log.Tracef("Pruning decayed failure report for vertex %v "+
"from Mission Control", vertex)
delete(m.failedVertexes, vertex)
continue
}
vertexes[vertex] = struct{}{}
}
// We'll also do the same for edges, but use the edgeDecay this time
// rather than the decay for vertexes.
edges := make(map[uint64]struct{})
for edge, pruneTime := range m.failedEdges {
if now.Sub(pruneTime) >= edgeDecay {
log.Tracef("Pruning decayed failure report for edge %v "+
"from Mission Control", edge)
delete(m.failedEdges, edge)
continue
}
edges[edge] = struct{}{}
}
m.Unlock()
log.Debugf("Mission Control returning prune view of %v edges, %v "+
"vertexes", len(edges), len(vertexes))
return graphPruneView{
edges: edges,
vertexes: vertexes,
}
}
// paymentSession is used during an HTLC routings session to prune the local
// chain view in response to failures, and also report those failures back to
// missionControl. The snapshot copied for this session will only ever grow,
// and will now be pruned after a decay like the main view within mission
// control. We do this as we want to avoid the case where we continually try a
// bad edge or route multiple times in a session. This can lead to an infinite
// loop if payment attempts take long enough. An additional set of edges can
// also be provided to assist in reaching the payment's destination.
type paymentSession struct {
pruneViewSnapshot graphPruneView
additionalEdges map[Vertex][]*channeldb.ChannelEdgePolicy
bandwidthHints map[uint64]lnwire.MilliSatoshi
mc *missionControl
}
// NewPaymentSession creates a new payment session backed by the latest prune
// view from Mission Control. An optional set of routing hints can be provided
// in order to populate additional edges to explore when finding a path to the
// payment's destination.
func (m *missionControl) NewPaymentSession(routeHints [][]HopHint,
target *btcec.PublicKey) (*paymentSession, error) {
viewSnapshot := m.GraphPruneView()
edges := make(map[Vertex][]*channeldb.ChannelEdgePolicy)
// Traverse through all of the available hop hints and include them in
// our edges map, indexed by the public key of the channel's starting
// node.
for _, routeHint := range routeHints {
// If multiple hop hints are provided within a single route
// hint, we'll assume they must be chained together and sorted
// in forward order in order to reach the target successfully.
for i, hopHint := range routeHint {
// In order to determine the end node of this hint,
// we'll need to look at the next hint's start node. If
// we've reached the end of the hints list, we can
// assume we've reached the destination.
endNode := &channeldb.LightningNode{}
if i != len(routeHint)-1 {
endNode.AddPubKey(routeHint[i+1].NodeID)
} else {
endNode.AddPubKey(target)
}
// Finally, create the channel edge from the hop hint
// and add it to list of edges corresponding to the node
// at the start of the channel.
edge := &channeldb.ChannelEdgePolicy{
Node: endNode,
ChannelID: hopHint.ChannelID,
FeeBaseMSat: lnwire.MilliSatoshi(
hopHint.FeeBaseMSat,
),
FeeProportionalMillionths: lnwire.MilliSatoshi(
hopHint.FeeProportionalMillionths,
),
TimeLockDelta: hopHint.CLTVExpiryDelta,
}
v := NewVertex(hopHint.NodeID)
edges[v] = append(edges[v], edge)
}
}
// We'll also obtain a set of bandwidthHints from the lower layer for
// each of our outbound channels. This will allow the path finding to
// skip any links that aren't active or just don't have enough
// bandwidth to carry the payment.
sourceNode, err := m.graph.SourceNode()
if err != nil {
return nil, err
}
bandwidthHints, err := generateBandwidthHints(
sourceNode, m.queryBandwidth,
)
if err != nil {
return nil, err
}
return &paymentSession{
pruneViewSnapshot: viewSnapshot,
additionalEdges: edges,
bandwidthHints: bandwidthHints,
mc: m,
}, nil
}
// generateBandwidthHints is a helper function that's utilized the main
// findPath function in order to obtain hints from the lower layer w.r.t to the
// available bandwidth of edges on the network. Currently, we'll only obtain
// bandwidth hints for the edges we directly have open ourselves. Obtaining
// these hints allows us to reduce the number of extraneous attempts as we can
// skip channels that are inactive, or just don't have enough bandwidth to
// carry the payment.
func generateBandwidthHints(sourceNode *channeldb.LightningNode,
queryBandwidth func(*channeldb.ChannelEdgeInfo) lnwire.MilliSatoshi) (map[uint64]lnwire.MilliSatoshi, error) {
// First, we'll collect the set of outbound edges from the target
// source node.
var localChans []*channeldb.ChannelEdgeInfo
err := sourceNode.ForEachChannel(nil, func(tx *bolt.Tx,
edgeInfo *channeldb.ChannelEdgeInfo,
_, _ *channeldb.ChannelEdgePolicy) error {
localChans = append(localChans, edgeInfo)
return nil
})
if err != nil {
return nil, err
}
// Now that we have all of our outbound edges, we'll populate the set
// of bandwidth hints, querying the lower switch layer for the most up
// to date values.
bandwidthHints := make(map[uint64]lnwire.MilliSatoshi)
for _, localChan := range localChans {
bandwidthHints[localChan.ChannelID] = queryBandwidth(localChan)
}
return bandwidthHints, nil
}
// ReportVertexFailure adds a vertex to the graph prune view after a client
// reports a routing failure localized to the vertex. The time the vertex was
// added is noted, as it'll be pruned from the shared view after a period of
// vertexDecay. However, the vertex will remain pruned for the *local* session.
// This ensures we don't retry this vertex during the payment attempt.
func (p *paymentSession) ReportVertexFailure(v Vertex) {
log.Debugf("Reporting vertex %v failure to Mission Control", v)
// First, we'll add the failed vertex to our local prune view snapshot.
p.pruneViewSnapshot.vertexes[v] = struct{}{}
// With the vertex added, we'll now report back to the global prune
// view, with this new piece of information so it can be utilized for
// new payment sessions.
p.mc.Lock()
p.mc.failedVertexes[v] = time.Now()
p.mc.Unlock()
}
// ReportChannelFailure adds a channel to the graph prune view. The time the
// channel was added is noted, as it'll be pruned from the global view after a
// period of edgeDecay. However, the edge will remain pruned for the duration
// of the *local* session. This ensures that we don't flap by continually
// retrying an edge after its pruning has expired.
//
// TODO(roasbeef): also add value attempted to send and capacity of channel
func (p *paymentSession) ReportChannelFailure(e uint64) {
log.Debugf("Reporting edge %v failure to Mission Control", e)
// First, we'll add the failed edge to our local prune view snapshot.
p.pruneViewSnapshot.edges[e] = struct{}{}
// With the edge added, we'll now report back to the global prune view,
// with this new piece of information so it can be utilized for new
// payment sessions.
p.mc.Lock()
p.mc.failedEdges[e] = time.Now()
p.mc.Unlock()
}
// RequestRoute returns a route which is likely to be capable for successfully
// routing the specified HTLC payment to the target node. Initially the first
// set of paths returned from this method may encounter routing failure along
// the way, however as more payments are sent, mission control will start to
// build an up to date view of the network itself. With each payment a new area
// will be explored, which feeds into the recommendations made for routing.
//
// NOTE: This function is safe for concurrent access.
func (p *paymentSession) RequestRoute(payment *LightningPayment,
height uint32, finalCltvDelta uint16) (*Route, error) {
// First, we'll obtain our current prune view snapshot. This view will
// only ever grow during the duration of this payment session, never
// shrinking.
pruneView := p.pruneViewSnapshot
log.Debugf("Mission Control session using prune view of %v "+
"edges, %v vertexes", len(pruneView.edges),
len(pruneView.vertexes))
// TODO(roasbeef): sync logic amongst dist sys
// Taking into account this prune view, we'll attempt to locate a path
// to our destination, respecting the recommendations from
// missionControl.
path, err := findPath(
nil, p.mc.graph, p.additionalEdges, p.mc.selfNode,
payment.Target, pruneView.vertexes, pruneView.edges,
payment.Amount, p.bandwidthHints,
)
if err != nil {
return nil, err
}
// With the next candidate path found, we'll attempt to turn this into
// a route by applying the time-lock and fee requirements.
sourceVertex := Vertex(p.mc.selfNode.PubKeyBytes)
route, err := newRoute(payment.Amount, sourceVertex, path, height,
finalCltvDelta)
if err != nil {
// TODO(roasbeef): return which edge/vertex didn't work
// out
return nil, err
}
return route, err
}
// ResetHistory resets the history of missionControl returning it to a state as
// if no payment attempts have been made.
func (m *missionControl) ResetHistory() {
m.Lock()
m.failedEdges = make(map[uint64]time.Time)
m.failedVertexes = make(map[Vertex]time.Time)
m.Unlock()
}