go-ethereum/swarm/network/stream/delivery_test.go
Balint Gabor 167159bc3c
Integrate encryption into dpa and chunker (#327)
* swarm: Integrate encryption into dpa and chunker

This is not a fully functional change.
Chunker api changed: it uses Putter and Getter interfaces to put and get
chunks from the store and to do the encryption/decryption if needed. These
putters and getters should also the hashing, so hashing related code
could be removed from chunker.
Some things are temporary, pyramid chunker doesn't work yet (only tree
chunker), and hashing and putting in dpa is not parallelized.

* swarm/storage: Use TreeChunker in DPA

PyramydChunker doesn't work yet with Putter/Getter

* swarm/storage: HasherStore needs a new hasher instance on every chunk
creation

This fixes the tests with pyramid chunker

* swarm/storage: Add back tree/pyramid comparison in chunker test

* swarm: Refactor chunker, new chunker instance is needed for each job

Chunker code is much simpler if all data of the job is stored on the
chunker object itself.

* swarm/storage: Remove commented code

* swarm/storage: Added comments

* cmd/swarm: Remove Branches from config

* swarm/storage: Remove commented code

* swarm/network: Remove chunker from syncer

* swarm/storage: Remove Chunker/Splitter/Joiner interfaces

They are not used anymore

* swarm/storage: Fix comment

* swarm/storage: Reenable chunker_test.TestDataAppend

* swarm/storage: Fix comments

* swarm/storage, cmd/swarm: Add FakeChunkStore

Instead of using a nil ChunkStore in hasherStore when we don't want the
hasherStore to actually store the chunk data, now a FakeChunkStore
instance should be used. Because of this the nil checks of the hasherStore.store
could be removed.

* swarm/storage: Use DefaultChunkSize az constant, not DefaultBranches

The branches what is changing, and it is dependent on chunk size and
hash size

* swarm/storage: Remove unnecessary return statement

* swarm/storage: Rename unfinishedChunk to unfinishedChunkData

* swarm/storage: Move NewTreeSplitterParams to chunker.go

It was in pyramid.go, although TreeSplitter is in chunker.go

* swarm/storage: Minor cleanup

incrementWorkerCount can be moved into runWorker
remove obsolete comment

* swarm/storage: Add TODO comment to chunker depth parameter

Depth can only be 0 because of a bug: https://github.com/ethersphere/go-ethereum/issues/344

* swarm/storage: Fix padding in hasherStore

There was a bug that data was not lengthened to padding

* swarm/storage: Add unit test for hasherStore

* swarm/storage: Decrease initial counter in span encryption

* swarm/storage: Renames in benchmark tests

* swarm/storage: Fixed pyramid benchmark test

* swarm/storage: Convert TestHasherStore to table-driven test

* swarm/storage, cmd/swarm: New MapChunkStore, simple ChunkStore implementation

It is not just the test which needs this kind of implementation

* swarm/storage: Cleanup on hashSize/chunkSize handling
2018-03-30 12:40:43 +02:00

676 lines
17 KiB
Go

// Copyright 2018 The go-ethereum Authors
// This file is part of the go-ethereum library.
//
// The go-ethereum library is free software: you can redistribute it and/or modify
// it under the terms of the GNU Lesser General Public License as published by
// the Free Software Foundation, either version 3 of the License, or
// (at your option) any later version.
//
// The go-ethereum library is distributed in the hope that it will be useful,
// but WITHOUT ANY WARRANTY; without even the implied warranty of
// MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
// GNU Lesser General Public License for more details.
//
// You should have received a copy of the GNU Lesser General Public License
// along with the go-ethereum library. If not, see <http://www.gnu.org/licenses/>.
package stream
import (
"bytes"
"context"
crand "crypto/rand"
"fmt"
"io"
"testing"
"time"
"github.com/ethereum/go-ethereum/common"
"github.com/ethereum/go-ethereum/log"
"github.com/ethereum/go-ethereum/p2p/discover"
"github.com/ethereum/go-ethereum/p2p/simulations"
p2ptest "github.com/ethereum/go-ethereum/p2p/testing"
"github.com/ethereum/go-ethereum/rpc"
"github.com/ethereum/go-ethereum/swarm/network"
streamTesting "github.com/ethereum/go-ethereum/swarm/network/stream/testing"
"github.com/ethereum/go-ethereum/swarm/storage"
)
var (
deliveries map[discover.NodeID]*Delivery
stores map[discover.NodeID]storage.ChunkStore
toAddr func(discover.NodeID) *network.BzzAddr
peerCount func(discover.NodeID) int
)
func TestStreamerRetrieveRequest(t *testing.T) {
tester, streamer, _, teardown, err := newStreamerTester(t)
defer teardown()
if err != nil {
t.Fatal(err)
}
peerID := tester.IDs[0]
streamer.delivery.RequestFromPeers(hash0[:], true)
err = tester.TestExchanges(p2ptest.Exchange{
Label: "RetrieveRequestMsg",
Expects: []p2ptest.Expect{
{
Code: 5,
Msg: &RetrieveRequestMsg{
Key: hash0[:],
SkipCheck: true,
},
Peer: peerID,
},
},
})
if err != nil {
t.Fatalf("Expected no error, got %v", err)
}
}
func TestStreamerUpstreamRetrieveRequestMsgExchangeWithoutStore(t *testing.T) {
tester, streamer, _, teardown, err := newStreamerTester(t)
defer teardown()
if err != nil {
t.Fatal(err)
}
peerID := tester.IDs[0]
chunk := storage.NewChunk(storage.Key(hash0[:]), nil)
peer := streamer.getPeer(peerID)
peer.handleSubscribeMsg(&SubscribeMsg{
Stream: NewStream(swarmChunkServerStreamName, "", false),
History: NewRange(0, 0),
Priority: Top,
})
err = tester.TestExchanges(p2ptest.Exchange{
Label: "RetrieveRequestMsg",
Triggers: []p2ptest.Trigger{
{
Code: 5,
Msg: &RetrieveRequestMsg{
Key: chunk.Key[:],
},
Peer: peerID,
},
},
Expects: []p2ptest.Expect{
{
Code: 1,
Msg: &OfferedHashesMsg{
HandoverProof: nil,
Hashes: nil,
From: 0,
To: 0,
},
Peer: peerID,
},
},
})
expectedError := `exchange #0 "RetrieveRequestMsg": timed out`
if err == nil || err.Error() != expectedError {
t.Fatalf("Expected error %v, got %v", expectedError, err)
}
}
// upstream request server receives a retrieve Request and responds with
// offered hashes or delivery if skipHash is set to true
func TestStreamerUpstreamRetrieveRequestMsgExchange(t *testing.T) {
tester, streamer, localStore, teardown, err := newStreamerTester(t)
defer teardown()
if err != nil {
t.Fatal(err)
}
peerID := tester.IDs[0]
peer := streamer.getPeer(peerID)
peer.handleSubscribeMsg(&SubscribeMsg{
Stream: NewStream(swarmChunkServerStreamName, "", false),
History: NewRange(0, 0),
Priority: Top,
})
hash := storage.Key(hash0[:])
chunk := storage.NewChunk(hash, nil)
chunk.SData = hash
localStore.Put(chunk)
chunk.WaitToStore()
err = tester.TestExchanges(p2ptest.Exchange{
Label: "RetrieveRequestMsg",
Triggers: []p2ptest.Trigger{
{
Code: 5,
Msg: &RetrieveRequestMsg{
Key: hash,
},
Peer: peerID,
},
},
Expects: []p2ptest.Expect{
{
Code: 1,
Msg: &OfferedHashesMsg{
HandoverProof: &HandoverProof{
Handover: &Handover{},
},
Hashes: hash,
From: 0,
// TODO: why is this 32???
To: 32,
Stream: NewStream(swarmChunkServerStreamName, "", false),
},
Peer: peerID,
},
},
})
if err != nil {
t.Fatal(err)
}
hash = storage.Key(hash1[:])
chunk = storage.NewChunk(hash, nil)
chunk.SData = hash1[:]
localStore.Put(chunk)
chunk.WaitToStore()
err = tester.TestExchanges(p2ptest.Exchange{
Label: "RetrieveRequestMsg",
Triggers: []p2ptest.Trigger{
{
Code: 5,
Msg: &RetrieveRequestMsg{
Key: hash,
SkipCheck: true,
},
Peer: peerID,
},
},
Expects: []p2ptest.Expect{
{
Code: 6,
Msg: &ChunkDeliveryMsg{
Key: hash,
SData: hash,
},
Peer: peerID,
},
},
})
if err != nil {
t.Fatal(err)
}
}
func TestStreamerDownstreamChunkDeliveryMsgExchange(t *testing.T) {
tester, streamer, localStore, teardown, err := newStreamerTester(t)
defer teardown()
if err != nil {
t.Fatal(err)
}
streamer.RegisterClientFunc("foo", func(p *Peer, t string, live bool) (Client, error) {
return &testClient{
t: t,
}, nil
})
peerID := tester.IDs[0]
stream := NewStream("foo", "", true)
err = streamer.Subscribe(peerID, stream, NewRange(5, 8), Top)
if err != nil {
t.Fatalf("Expected no error, got %v", err)
}
chunkKey := hash0[:]
chunkData := hash1[:]
chunk, created := localStore.GetOrCreateRequest(chunkKey)
if !created {
t.Fatal("chunk already exists")
}
select {
case <-chunk.ReqC:
t.Fatal("chunk is already received")
default:
}
err = tester.TestExchanges(p2ptest.Exchange{
Label: "Subscribe message",
Expects: []p2ptest.Expect{
{
Code: 4,
Msg: &SubscribeMsg{
Stream: stream,
History: NewRange(5, 8),
Priority: Top,
},
Peer: peerID,
},
},
},
p2ptest.Exchange{
Label: "ChunkDeliveryRequest message",
Triggers: []p2ptest.Trigger{
{
Code: 6,
Msg: &ChunkDeliveryMsg{
Key: chunkKey,
SData: chunkData,
},
Peer: peerID,
},
},
})
if err != nil {
t.Fatalf("Expected no error, got %v", err)
}
timeout := time.NewTimer(1 * time.Second)
select {
case <-timeout.C:
t.Fatal("timeout receiving chunk")
case <-chunk.ReqC:
}
storedChunk, err := localStore.Get(chunkKey)
if err != nil {
t.Fatalf("Expected no error, got %v", err)
}
if !bytes.Equal(storedChunk.SData, chunkData) {
t.Fatal("Retrieved chunk has different data than original")
}
}
func TestDeliveryFromNodes(t *testing.T) {
testDeliveryFromNodes(t, 2, 1, dataChunkCount, true)
testDeliveryFromNodes(t, 2, 1, dataChunkCount, false)
testDeliveryFromNodes(t, 4, 1, dataChunkCount, true)
testDeliveryFromNodes(t, 4, 1, dataChunkCount, false)
testDeliveryFromNodes(t, 8, 1, dataChunkCount, true)
testDeliveryFromNodes(t, 8, 1, dataChunkCount, false)
testDeliveryFromNodes(t, 16, 1, dataChunkCount, true)
testDeliveryFromNodes(t, 16, 1, dataChunkCount, false)
}
func testDeliveryFromNodes(t *testing.T, nodes, conns, chunkCount int, skipCheck bool) {
defaultSkipCheck = skipCheck
toAddr = network.NewAddrFromNodeID
conf := &streamTesting.RunConfig{
Adapter: *adapter,
NodeCount: nodes,
ConnLevel: conns,
ToAddr: toAddr,
Services: services,
EnableMsgEvents: false,
}
sim, teardown, err := streamTesting.NewSimulation(conf)
defer teardown()
if err != nil {
t.Fatal(err.Error())
}
stores = make(map[discover.NodeID]storage.ChunkStore)
deliveries = make(map[discover.NodeID]*Delivery)
for i, id := range sim.IDs {
stores[id] = sim.Stores[i]
}
peerCount = func(id discover.NodeID) int {
if sim.IDs[0] == id || sim.IDs[nodes-1] == id {
return 1
}
return 2
}
// here we distribute chunks of a random file into Stores of nodes 1 to nodes
rrdpa := storage.NewDPA(newRoundRobinStore(sim.Stores[1:]...), storage.NewDPAParams())
size := chunkCount * chunkSize
fileHash, wait, err := rrdpa.Store(io.LimitReader(crand.Reader, int64(size)), int64(size), false)
// wait until all chunks stored
wait()
if err != nil {
t.Fatal(err.Error())
}
errc := make(chan error, 1)
waitPeerErrC = make(chan error)
quitC := make(chan struct{})
action := func(ctx context.Context) error {
// each node Subscribes to each other's swarmChunkServerStreamName
// need to wait till an aynchronous process registers the peers in streamer.peers
// that is used by Subscribe
// using a global err channel to share betweem action and node service
i := 0
for err := range waitPeerErrC {
if err != nil {
return fmt.Errorf("error waiting for peers: %s", err)
}
i++
if i == nodes {
break
}
}
// each node subscribes to the upstream swarm chunk server stream
// which responds to chunk retrieve requests all but the last node in the chain does not
for j := 0; j < nodes-1; j++ {
id := sim.IDs[j]
err := sim.CallClient(id, func(client *rpc.Client) error {
err := streamTesting.WatchDisconnections(id, client, errc, quitC)
if err != nil {
return err
}
ctx, cancel := context.WithTimeout(ctx, 1*time.Second)
defer cancel()
sid := sim.IDs[j+1]
return client.CallContext(ctx, nil, "stream_subscribeStream", sid, NewStream(swarmChunkServerStreamName, "", false), NewRange(0, 0), Top)
})
if err != nil {
return err
}
}
// create a retriever dpa for the pivot node
delivery := deliveries[sim.IDs[0]]
retrieveFunc := func(chunk *storage.Chunk) error {
return delivery.RequestFromPeers(chunk.Key[:], skipCheck)
}
netStore := storage.NewNetStore(sim.Stores[0].(*storage.LocalStore), retrieveFunc)
dpa := storage.NewDPA(netStore, storage.NewDPAParams())
go func() {
// start the retrieval on the pivot node - this will spawn retrieve requests for missing chunks
// we must wait for the peer connections to have started before requesting
n, err := readAll(dpa, fileHash)
log.Info(fmt.Sprintf("retrieved %v", fileHash), "read", n, "err", err)
if err != nil {
errc <- fmt.Errorf("requesting chunks action error: %v", err)
}
}()
return nil
}
check := func(ctx context.Context, id discover.NodeID) (bool, error) {
select {
case err := <-errc:
return false, err
case <-ctx.Done():
return false, ctx.Err()
default:
}
var total int64
err := sim.CallClient(id, func(client *rpc.Client) error {
ctx, cancel := context.WithTimeout(ctx, 5*time.Second)
defer cancel()
return client.CallContext(ctx, &total, "stream_readAll", common.BytesToHash(fileHash))
})
log.Info(fmt.Sprintf("check if %08x is available locally: number of bytes read %v/%v (error: %v)", fileHash, total, size, err))
if err != nil || total != int64(size) {
return false, nil
}
return true, nil
}
conf.Step = &simulations.Step{
Action: action,
Trigger: streamTesting.Trigger(10*time.Millisecond, quitC, sim.IDs[0]),
// we are only testing the pivot node (net.Nodes[0])
Expect: &simulations.Expectation{
Nodes: sim.IDs[0:1],
Check: check,
},
}
startedAt := time.Now()
timeout := 300 * time.Second
ctx, cancel := context.WithTimeout(context.Background(), timeout)
defer cancel()
result, err := sim.Run(ctx, conf)
finishedAt := time.Now()
if err != nil {
t.Fatalf("Setting up simulation failed: %v", err)
}
if result.Error != nil {
t.Fatalf("Simulation failed: %s", result.Error)
}
streamTesting.CheckResult(t, result, startedAt, finishedAt)
}
func BenchmarkDeliveryFromNodesWithoutCheck(b *testing.B) {
for chunks := 32; chunks <= 128; chunks *= 2 {
for i := 2; i < 32; i *= 2 {
b.Run(
fmt.Sprintf("nodes=%v,chunks=%v", i, chunks),
func(b *testing.B) {
benchmarkDeliveryFromNodes(b, i, 1, chunks, true)
},
)
}
}
}
func BenchmarkDeliveryFromNodesWithCheck(b *testing.B) {
for chunks := 32; chunks <= 128; chunks *= 2 {
for i := 2; i < 32; i *= 2 {
b.Run(
fmt.Sprintf("nodes=%v,chunks=%v", i, chunks),
func(b *testing.B) {
benchmarkDeliveryFromNodes(b, i, 1, chunks, false)
},
)
}
}
}
func benchmarkDeliveryFromNodes(b *testing.B, nodes, conns, chunkCount int, skipCheck bool) {
defaultSkipCheck = skipCheck
toAddr = network.NewAddrFromNodeID
timeout := 300 * time.Second
ctx, cancel := context.WithTimeout(context.Background(), timeout)
defer cancel()
conf := &streamTesting.RunConfig{
Adapter: *adapter,
NodeCount: nodes,
ConnLevel: conns,
ToAddr: toAddr,
Services: services,
EnableMsgEvents: false,
}
sim, teardown, err := streamTesting.NewSimulation(conf)
defer teardown()
if err != nil {
b.Fatal(err.Error())
}
stores = make(map[discover.NodeID]storage.ChunkStore)
deliveries = make(map[discover.NodeID]*Delivery)
for i, id := range sim.IDs {
stores[id] = sim.Stores[i]
}
peerCount = func(id discover.NodeID) int {
if sim.IDs[0] == id || sim.IDs[nodes-1] == id {
return 1
}
return 2
}
// wait channel for all nodes all peer connections to set up
waitPeerErrC = make(chan error)
// create a dpa for the last node in the chain which we are gonna write to
remoteDpa := storage.NewDPA(sim.Stores[nodes-1], storage.NewDPAParams())
// channel to signal simulation initialisation with action call complete
// or node disconnections
disconnectC := make(chan error)
quitC := make(chan struct{})
initC := make(chan error)
action := func(ctx context.Context) error {
// each node Subscribes to each other's swarmChunkServerStreamName
// need to wait till an aynchronous process registers the peers in streamer.peers
// that is used by Subscribe
// waitPeerErrC using a global err channel to share betweem action and node service
i := 0
for err := range waitPeerErrC {
if err != nil {
return fmt.Errorf("error waiting for peers: %s", err)
}
i++
if i == nodes {
break
}
}
var err error
// each node except the last one subscribes to the upstream swarm chunk server stream
// which responds to chunk retrieve requests
for j := 0; j < nodes-1; j++ {
id := sim.IDs[j]
err = sim.CallClient(id, func(client *rpc.Client) error {
err := streamTesting.WatchDisconnections(id, client, disconnectC, quitC)
if err != nil {
return err
}
ctx, cancel := context.WithTimeout(ctx, 1*time.Second)
defer cancel()
sid := sim.IDs[j+1] // the upstream peer's id
return client.CallContext(ctx, nil, "stream_subscribeStream", sid, NewStream(swarmChunkServerStreamName, "", false), NewRange(0, 0), Top)
})
if err != nil {
break
}
}
initC <- err
return nil
}
// the check function is only triggered when the benchmark finishes
trigger := make(chan discover.NodeID)
check := func(ctx context.Context, id discover.NodeID) (_ bool, err error) {
return true, nil
}
conf.Step = &simulations.Step{
Action: action,
Trigger: trigger,
// we are only testing the pivot node (net.Nodes[0])
Expect: &simulations.Expectation{
Nodes: sim.IDs[0:1],
Check: check,
},
}
// run the simulation in the background
errc := make(chan error)
go func() {
_, err := sim.Run(ctx, conf)
close(quitC)
errc <- err
}()
// wait for simulation action to complete stream subscriptions
err = <-initC
if err != nil {
b.Fatalf("simulation failed to initialise. expected no error. got %v", err)
}
// create a retriever dpa for the pivot node
// by now deliveries are set for each node by the streamer service
delivery := deliveries[sim.IDs[0]]
retrieveFunc := func(chunk *storage.Chunk) error {
return delivery.RequestFromPeers(chunk.Key[:], skipCheck)
}
netStore := storage.NewNetStore(sim.Stores[0].(*storage.LocalStore), retrieveFunc)
// benchmark loop
b.ResetTimer()
b.StopTimer()
Loop:
for i := 0; i < b.N; i++ {
// uploading chunkCount random chunks to the last node
hashes := make([]storage.Key, chunkCount)
for i := 0; i < chunkCount; i++ {
// create actual size real chunks
hash, wait, err := remoteDpa.Store(io.LimitReader(crand.Reader, int64(chunkSize)), int64(chunkSize), false)
// wait until all chunks stored
wait()
if err != nil {
b.Fatalf("expected no error. got %v", err)
}
// collect the hashes
hashes[i] = hash
}
// now benchmark the actual retrieval
// netstore.Get is called for each hash in a go routine and errors are collected
b.StartTimer()
errs := make(chan error)
for _, hash := range hashes {
go func(h storage.Key) {
_, err := netStore.Get(h)
log.Warn("test check netstore get", "hash", h, "err", err)
errs <- err
}(hash)
}
// count and report retrieval errors
// if there are misses then chunk timeout is too low for the distance and volume (?)
var total, misses int
for err := range errs {
if err != nil {
log.Warn(err.Error())
misses++
}
total++
if total == chunkCount {
break
}
}
b.StopTimer()
select {
case err = <-disconnectC:
if err != nil {
break Loop
}
default:
}
if misses > 0 {
err = fmt.Errorf("%v chunk not found out of %v", misses, total)
break Loop
}
}
select {
case <-quitC:
case trigger <- sim.IDs[0]:
}
if err == nil {
err = <-errc
} else {
if e := <-errc; e != nil {
b.Errorf("sim.Run function error: %v", e)
}
}
// benchmark over, trigger the check function to conclude the simulation
if err != nil {
b.Fatalf("expected no error. got %v", err)
}
}