Skip to content
This repository has been archived by the owner on Feb 1, 2023. It is now read-only.

Commit

Permalink
Merge pull request #276 from ipfs/fix/prune-dont-have
Browse files Browse the repository at this point in the history
Be less aggressive when pruning peers from session
  • Loading branch information
Stebalien authored Mar 6, 2020
2 parents 159748c + 568a984 commit 418d88c
Show file tree
Hide file tree
Showing 8 changed files with 126 additions and 52 deletions.
2 changes: 1 addition & 1 deletion benchmarks_test.go
Original file line number Diff line number Diff line change
Expand Up @@ -130,7 +130,7 @@ var mixedBenches = []mixedBench{
mixedBench{bench{"3Nodes-Overlap3-OneAtATime", 3, 10, overlap2, oneAtATime}, 1, 2},
mixedBench{bench{"3Nodes-AllToAll-OneAtATime", 3, 10, allToAll, oneAtATime}, 1, 2},
mixedBench{bench{"3Nodes-Overlap3-AllConcurrent", 3, 10, overlap2, fetchAllConcurrent}, 1, 2},
mixedBench{bench{"3Nodes-Overlap3-UnixfsFetch", 3, 100, overlap2, unixfsFileFetch}, 1, 2},
// mixedBench{bench{"3Nodes-Overlap3-UnixfsFetch", 3, 100, overlap2, unixfsFileFetch}, 1, 2},
}

func BenchmarkFetchFromOldBitswap(b *testing.B) {
Expand Down
8 changes: 4 additions & 4 deletions internal/decision/engine_test.go
Original file line number Diff line number Diff line change
Expand Up @@ -1007,9 +1007,9 @@ func TestTaggingPeers(t *testing.T) {
}

func TestTaggingUseful(t *testing.T) {
peerSampleInterval := 2 * time.Millisecond
peerSampleInterval := 5 * time.Millisecond

ctx, cancel := context.WithTimeout(context.Background(), 1*time.Second)
ctx, cancel := context.WithTimeout(context.Background(), 2*time.Second)
defer cancel()
me := newTestEngine(ctx, "engine", peerSampleInterval)
friend := peer.ID("friend")
Expand All @@ -1023,11 +1023,11 @@ func TestTaggingUseful(t *testing.T) {
t.Fatal("Peers should be untagged but weren't")
}
me.Engine.MessageSent(friend, msg)
time.Sleep(peerSampleInterval * 2)
time.Sleep(8 * time.Millisecond)
if me.PeerTagger.count(me.Engine.tagUseful) != 1 {
t.Fatal("Peers should be tagged but weren't")
}
time.Sleep(peerSampleInterval * 8)
time.Sleep(peerSampleInterval * 10)
}

if me.PeerTagger.count(me.Engine.tagUseful) == 0 {
Expand Down
1 change: 1 addition & 0 deletions internal/messagequeue/messagequeue.go
Original file line number Diff line number Diff line change
Expand Up @@ -46,6 +46,7 @@ type MessageNetwork interface {
NewMessageSender(context.Context, peer.ID) (bsnet.MessageSender, error)
Latency(peer.ID) time.Duration
Ping(context.Context, peer.ID) ping.Result
Self() peer.ID
}

// MessageQueue implements queue of want messages to send to peers.
Expand Down
4 changes: 2 additions & 2 deletions internal/session/session.go
Original file line number Diff line number Diff line change
Expand Up @@ -4,9 +4,9 @@ import (
"context"
"time"

// lu "github.com/ipfs/go-bitswap/internal/logutil"
bsbpm "github.com/ipfs/go-bitswap/internal/blockpresencemanager"
bsgetter "github.com/ipfs/go-bitswap/internal/getter"
lu "github.com/ipfs/go-bitswap/internal/logutil"
notifications "github.com/ipfs/go-bitswap/internal/notifications"
bspm "github.com/ipfs/go-bitswap/internal/peermanager"
bssim "github.com/ipfs/go-bitswap/internal/sessioninterestmanager"
Expand Down Expand Up @@ -340,7 +340,7 @@ func (s *Session) broadcastWantHaves(ctx context.Context, wants []cid.Cid) {
// Search for providers who have the first want in the list.
// Typically if the provider has the first block they will have
// the rest of the blocks also.
log.Warnf("Ses%d: FindMorePeers with want 0 of %d wants", s.id, len(wants))
log.Warnf("Ses%d: FindMorePeers with want %s (1st of %d wants)", s.id, lu.C(wants[0]), len(wants))
s.findMorePeers(ctx, wants[0])
}
s.resetIdleTick()
Expand Down
8 changes: 5 additions & 3 deletions internal/session/sessionwants.go
Original file line number Diff line number Diff line change
Expand Up @@ -56,7 +56,7 @@ func (sw *sessionWants) GetNextWants(limit int) []cid.Cid {
func (sw *sessionWants) WantsSent(ks []cid.Cid) {
now := time.Now()
for _, c := range ks {
if _, ok := sw.liveWants[c]; !ok {
if _, ok := sw.liveWants[c]; !ok && sw.toFetch.Has(c) {
sw.toFetch.Remove(c)
sw.liveWants[c] = now
}
Expand All @@ -83,8 +83,7 @@ func (sw *sessionWants) BlocksReceived(ks []cid.Cid) ([]cid.Cid, time.Duration)
totalLatency += now.Sub(sentAt)
}

// Remove the CID from the live wants / toFetch queue and add it
// to the past wants
// Remove the CID from the live wants / toFetch queue
delete(sw.liveWants, c)
sw.toFetch.Remove(c)
}
Expand All @@ -96,6 +95,9 @@ func (sw *sessionWants) BlocksReceived(ks []cid.Cid) ([]cid.Cid, time.Duration)
// PrepareBroadcast saves the current time for each live want and returns the
// live want CIDs.
func (sw *sessionWants) PrepareBroadcast() []cid.Cid {
// TODO: Change this to return wants in order so that the session will
// send out Find Providers request for the first want
// (Note that maps return keys in random order)
now := time.Now()
live := make([]cid.Cid, 0, len(sw.liveWants))
for c := range sw.liveWants {
Expand Down
83 changes: 55 additions & 28 deletions internal/session/sessionwantsender.go
Original file line number Diff line number Diff line change
Expand Up @@ -4,6 +4,7 @@ import (
"context"

bsbpm "github.com/ipfs/go-bitswap/internal/blockpresencemanager"
lu "github.com/ipfs/go-bitswap/internal/logutil"

cid "github.com/ipfs/go-cid"
peer "github.com/libp2p/go-libp2p-core/peer"
Expand Down Expand Up @@ -298,16 +299,41 @@ func (sws *sessionWantSender) trackWant(c cid.Cid) {
// processUpdates processes incoming blocks and HAVE / DONT_HAVEs.
// It returns all DONT_HAVEs.
func (sws *sessionWantSender) processUpdates(updates []update) []cid.Cid {
prunePeers := make(map[peer.ID]struct{})
dontHaves := cid.NewSet()
// Process received blocks keys
blkCids := cid.NewSet()
for _, upd := range updates {
// TODO: If there is a timeout for the want from the peer, remove want.sentTo
// so the want can be sent to another peer (and blacklist the peer?)
// TODO: If a peer is no longer available, check if all providers of
// each CID have been exhausted
for _, c := range upd.ks {
blkCids.Add(c)
log.Warnf("received block %s", lu.C(c))
// Remove the want
removed := sws.removeWant(c)
if removed != nil {
// Inform the peer tracker that this peer was the first to send
// us the block
sws.peerRspTrkr.receivedBlockFrom(upd.from)
}
delete(sws.peerConsecutiveDontHaves, upd.from)
}
}

// For each DONT_HAVE
// Process received DONT_HAVEs
dontHaves := cid.NewSet()
prunePeers := make(map[peer.ID]struct{})
for _, upd := range updates {
for _, c := range upd.dontHaves {
// Track the number of consecutive DONT_HAVEs each peer receives
if sws.peerConsecutiveDontHaves[upd.from] == peerDontHaveLimit {
prunePeers[upd.from] = struct{}{}
} else {
sws.peerConsecutiveDontHaves[upd.from]++
}

// If we already received a block for the want, there's no need to
// update block presence etc
if blkCids.Has(c) {
continue
}

dontHaves.Add(c)

// Update the block presence for the peer
Expand All @@ -322,40 +348,41 @@ func (sws *sessionWantSender) processUpdates(updates []update) []cid.Cid {
sws.setWantSentTo(c, "")
}
}

// Track the number of consecutive DONT_HAVEs each peer receives
if sws.peerConsecutiveDontHaves[upd.from] == peerDontHaveLimit {
prunePeers[upd.from] = struct{}{}
} else {
sws.peerConsecutiveDontHaves[upd.from]++
}
}
}

// For each HAVE
// Process received HAVEs
for _, upd := range updates {
for _, c := range upd.haves {
// Update the block presence for the peer
sws.updateWantBlockPresence(c, upd.from)
delete(sws.peerConsecutiveDontHaves, upd.from)
}

// For each received block
for _, c := range upd.ks {
// Remove the want
removed := sws.removeWant(c)
if removed != nil {
// Inform the peer tracker that this peer was the first to send
// us the block
sws.peerRspTrkr.receivedBlockFrom(upd.from)
// If we haven't already received a block for the want
if !blkCids.Has(c) {
// Update the block presence for the peer
sws.updateWantBlockPresence(c, upd.from)
}

// Clear the consecutive DONT_HAVE count for the peer
delete(sws.peerConsecutiveDontHaves, upd.from)
delete(prunePeers, upd.from)
}
}

// If any peers have sent us too many consecutive DONT_HAVEs, remove them
// from the session
for p := range prunePeers {
// Before removing the peer from the session, check if the peer
// sent us a HAVE for a block that we want
for c := range sws.wants {
if sws.bpm.PeerHasBlock(p, c) {
delete(prunePeers, p)
break
}
}
}
if len(prunePeers) > 0 {
go func() {
for p := range prunePeers {
// Peer doesn't have anything we want, so remove it
log.Infof("peer %s sent too many dont haves", lu.P(p))
sws.SignalAvailability(p, false)
}
}()
Expand Down
67 changes: 54 additions & 13 deletions internal/session/sessionwantsender_test.go
Original file line number Diff line number Diff line change
Expand Up @@ -529,9 +529,8 @@ func TestConsecutiveDontHaveLimit(t *testing.T) {
// Add all cids as wants
spm.Add(cids)

// Receive a HAVE from peer (adds it to the session)
bpm.ReceiveFrom(p, cids[:1], []cid.Cid{})
spm.Update(p, []cid.Cid{}, cids[:1], []cid.Cid{})
// Receive a block from peer (adds it to the session)
spm.Update(p, cids[:1], []cid.Cid{}, []cid.Cid{})

// Wait for processing to complete
time.Sleep(10 * time.Millisecond)
Expand Down Expand Up @@ -586,9 +585,8 @@ func TestConsecutiveDontHaveLimitInterrupted(t *testing.T) {
// Add all cids as wants
spm.Add(cids)

// Receive a HAVE from peer (adds it to the session)
bpm.ReceiveFrom(p, cids[:1], []cid.Cid{})
spm.Update(p, []cid.Cid{}, cids[:1], []cid.Cid{})
// Receive a block from peer (adds it to the session)
spm.Update(p, cids[:1], []cid.Cid{}, []cid.Cid{})

// Wait for processing to complete
time.Sleep(5 * time.Millisecond)
Expand Down Expand Up @@ -642,9 +640,8 @@ func TestConsecutiveDontHaveReinstateAfterRemoval(t *testing.T) {
// Add all cids as wants
spm.Add(cids)

// Receive a HAVE from peer (adds it to the session)
bpm.ReceiveFrom(p, cids[:1], []cid.Cid{})
spm.Update(p, []cid.Cid{}, cids[:1], []cid.Cid{})
// Receive a block from peer (adds it to the session)
spm.Update(p, cids[:1], []cid.Cid{}, []cid.Cid{})

// Wait for processing to complete
time.Sleep(5 * time.Millisecond)
Expand All @@ -661,7 +658,7 @@ func TestConsecutiveDontHaveReinstateAfterRemoval(t *testing.T) {
}

// Wait for processing to complete
time.Sleep(5 * time.Millisecond)
time.Sleep(10 * time.Millisecond)

// Session should remove peer
if has := fpm.HasPeer(p); has {
Expand All @@ -673,7 +670,7 @@ func TestConsecutiveDontHaveReinstateAfterRemoval(t *testing.T) {
spm.Update(p, []cid.Cid{}, cids[:1], []cid.Cid{})

// Wait for processing to complete
time.Sleep(5 * time.Millisecond)
time.Sleep(10 * time.Millisecond)

// Peer should be available
if has := fpm.HasPeer(p); !has {
Expand All @@ -689,7 +686,7 @@ func TestConsecutiveDontHaveReinstateAfterRemoval(t *testing.T) {
}

// Wait for processing to complete
time.Sleep(5 * time.Millisecond)
time.Sleep(10 * time.Millisecond)

// Peer should be available
if has := fpm.HasPeer(p); !has {
Expand All @@ -703,10 +700,54 @@ func TestConsecutiveDontHaveReinstateAfterRemoval(t *testing.T) {
}

// Wait for processing to complete
time.Sleep(5 * time.Millisecond)
time.Sleep(10 * time.Millisecond)

// Session should remove peer
if has := fpm.HasPeer(p); has {
t.Fatal("Expected peer not to be available")
}
}

func TestConsecutiveDontHaveDontRemoveIfHasWantedBlock(t *testing.T) {
cids := testutil.GenerateCids(peerDontHaveLimit + 10)
p := testutil.GeneratePeers(1)[0]
sid := uint64(1)
pm := newMockPeerManager()
fpm := newFakeSessionPeerManager()
bpm := bsbpm.New()
onSend := func(peer.ID, []cid.Cid, []cid.Cid) {}
onPeersExhausted := func([]cid.Cid) {}
spm := newSessionWantSender(context.Background(), sid, pm, fpm, bpm, onSend, onPeersExhausted)

go spm.Run()

// Add all cids as wants
spm.Add(cids)

// Receive a HAVE from peer (adds it to the session)
bpm.ReceiveFrom(p, cids[:1], []cid.Cid{})
spm.Update(p, []cid.Cid{}, cids[:1], []cid.Cid{})

// Wait for processing to complete
time.Sleep(10 * time.Millisecond)

// Peer should be available
if has := fpm.HasPeer(p); !has {
t.Fatal("Expected peer to be available")
}

// Receive DONT_HAVEs from peer that exceed limit
for _, c := range cids[1 : peerDontHaveLimit+5] {
bpm.ReceiveFrom(p, []cid.Cid{}, []cid.Cid{c})
spm.Update(p, []cid.Cid{}, []cid.Cid{}, []cid.Cid{c})
}

// Wait for processing to complete
time.Sleep(20 * time.Millisecond)

// Peer should still be available because it has a block that we want.
// (We received a HAVE for cid 0 but didn't yet receive the block)
if has := fpm.HasPeer(p); !has {
t.Fatal("Expected peer to be available")
}
}
5 changes: 4 additions & 1 deletion internal/sessionpeermanager/sessionpeermanager.go
Original file line number Diff line number Diff line change
Expand Up @@ -4,6 +4,7 @@ import (
"fmt"
"sync"

lu "github.com/ipfs/go-bitswap/internal/logutil"
logging "github.com/ipfs/go-log"

peer "github.com/libp2p/go-libp2p-core/peer"
Expand Down Expand Up @@ -61,7 +62,7 @@ func (spm *SessionPeerManager) AddPeer(p peer.ID) bool {
// connection
spm.tagger.TagPeer(p, spm.tag, sessionPeerTagValue)

log.Infof("Added peer %s to session: %d peers\n", p, len(spm.peers))
log.Debugf("Added peer %s to session (%d peers)\n", p, len(spm.peers))
return true
}

Expand All @@ -77,6 +78,8 @@ func (spm *SessionPeerManager) RemovePeer(p peer.ID) bool {

delete(spm.peers, p)
spm.tagger.UntagPeer(p, spm.tag)

log.Debugf("Removed peer %s from session (%d peers)", lu.P(p), len(spm.peers))
return true
}

Expand Down

0 comments on commit 418d88c

Please sign in to comment.