go-ethereum/swarm/network/stream/snapshot_retrieval_test.go
Ferenc Szabo e38b227ce6 Ci race detector handle failing tests (#19143)
* swarm/storage: increase mget timeout in common_test.go

 TestDbStoreCorrect_1k sometimes timed out with -race on Travis.

--- FAIL: TestDbStoreCorrect_1k (24.63s)
    common_test.go:194: testStore failed: timed out after 10s

* swarm: remove unused vars from TestSnapshotSyncWithServer

nodeCount and chunkCount is returned from setupSim and those values
we use.

* swarm: move race/norace helpers from stream to testutil

As we will need to use the flag in other packages, too.

* swarm: refactor TestSwarmNetwork case

Extract long running test cases for better visibility.

* swarm/network: skip TestSyncingViaGlobalSync with -race

As panics on Travis.

panic: runtime error: invalid memory address or nil pointer dereference
[signal SIGSEGV: segmentation violation code=0x1 addr=0x0 pc=0x7e351b]

* swarm: run TestSwarmNetwork with fewer nodes with -race

As otherwise we always get test failure with `network_test.go:374:
context deadline exceeded` even with raised `Timeout`.

* swarm/network: run TestDeliveryFromNodes with fewer nodes with -race

Test on Travis times out with 8 or more nodes if -race flag is present.

* swarm/network: smaller node count for discovery tests with -race

TestDiscoveryPersistenceSimulationSimAdapters failed on Travis with
`-race` flag present. The failure was due to extensive memory usage,
coming from the CGO runtime. Using a smaller node count resolves the
issue.

=== RUN   TestDiscoveryPersistenceSimulationSimAdapter
==7227==ERROR: ThreadSanitizer failed to allocate 0x80000 (524288) bytes of clock allocator (error code: 12)
FATAL: ThreadSanitizer CHECK failed: ./gotsan.cc:6976 "((0 && "unable to mmap")) != (0)" (0x0, 0x0)
FAIL    github.com/ethereum/go-ethereum/swarm/network/simulations/discovery     804.826s

* swarm/network: run TestFileRetrieval with fewer nodes with -race

Otherwise we get a failure due to extensive memory usage, as the CGO
runtime cannot allocate more bytes.

=== RUN   TestFileRetrieval
==7366==ERROR: ThreadSanitizer failed to allocate 0x80000 (524288) bytes of clock allocator (error code: 12)
FATAL: ThreadSanitizer CHECK failed: ./gotsan.cc:6976 "((0 && "unable to mmap")) != (0)" (0x0, 0x0)
FAIL	github.com/ethereum/go-ethereum/swarm/network/stream	155.165s

* swarm/network: run TestRetrieval with fewer nodes with -race

Otherwise we get a failure due to extensive memory usage, as the CGO
runtime cannot allocate more bytes ("ThreadSanitizer failed to
allocate").

* swarm/network: skip flaky TestGetSubscriptionsRPC on Travis w/ -race

Test fails a lot with something like:
 streamer_test.go:1332: Real subscriptions and expected amount don't match; real: 0, expected: 20

* swarm/storage: skip TestDB_SubscribePull* tests on Travis w/ -race

Travis just hangs...

ok  	github.com/ethereum/go-ethereum/swarm/storage/feed/lookup	1.307s
keepalive
keepalive
keepalive

or panics after a while.

Without these tests the race detector job is now stable. Let's
invetigate these tests in a separate issue:
https://github.com/ethersphere/go-ethereum/issues/1245
2019-02-20 22:57:42 +01:00

314 lines
9.3 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 (
"context"
"fmt"
"sync"
"testing"
"time"
"github.com/ethereum/go-ethereum/swarm/testutil"
"github.com/ethereum/go-ethereum/node"
"github.com/ethereum/go-ethereum/p2p/enode"
"github.com/ethereum/go-ethereum/p2p/simulations/adapters"
"github.com/ethereum/go-ethereum/swarm/log"
"github.com/ethereum/go-ethereum/swarm/network/simulation"
"github.com/ethereum/go-ethereum/swarm/state"
"github.com/ethereum/go-ethereum/swarm/storage"
)
//constants for random file generation
const (
minFileSize = 2
maxFileSize = 40
)
//This test is a retrieval test for nodes.
//A configurable number of nodes can be
//provided to the test.
//Files are uploaded to nodes, other nodes try to retrieve the file
//Number of nodes can be provided via commandline too.
func TestFileRetrieval(t *testing.T) {
var nodeCount []int
if *nodes != 0 {
nodeCount = []int{*nodes}
} else {
nodeCount = []int{16}
if *longrunning {
nodeCount = append(nodeCount, 32, 64, 128)
} else if testutil.RaceEnabled {
nodeCount = []int{4}
}
}
for _, nc := range nodeCount {
if err := runFileRetrievalTest(nc); err != nil {
t.Error(err)
}
}
}
//This test is a retrieval test for nodes.
//One node is randomly selected to be the pivot node.
//A configurable number of chunks and nodes can be
//provided to the test, the number of chunks is uploaded
//to the pivot node and other nodes try to retrieve the chunk(s).
//Number of chunks and nodes can be provided via commandline too.
func TestRetrieval(t *testing.T) {
//if nodes/chunks have been provided via commandline,
//run the tests with these values
if *nodes != 0 && *chunks != 0 {
err := runRetrievalTest(t, *chunks, *nodes)
if err != nil {
t.Fatal(err)
}
} else {
nodeCnt := []int{16}
chnkCnt := []int{32}
if *longrunning {
nodeCnt = []int{16, 32, 128}
chnkCnt = []int{4, 32, 256}
} else if testutil.RaceEnabled {
nodeCnt = []int{4}
chnkCnt = []int{4}
}
for _, n := range nodeCnt {
for _, c := range chnkCnt {
t.Run(fmt.Sprintf("TestRetrieval_%d_%d", n, c), func(t *testing.T) {
err := runRetrievalTest(t, c, n)
if err != nil {
t.Fatal(err)
}
})
}
}
}
}
var retrievalSimServiceMap = map[string]simulation.ServiceFunc{
"streamer": func(ctx *adapters.ServiceContext, bucket *sync.Map) (s node.Service, cleanup func(), err error) {
addr, netStore, delivery, clean, err := newNetStoreAndDelivery(ctx, bucket)
if err != nil {
return nil, nil, err
}
r := NewRegistry(addr.ID(), delivery, netStore, state.NewInmemoryStore(), &RegistryOptions{
Retrieval: RetrievalEnabled,
Syncing: SyncingAutoSubscribe,
SyncUpdateDelay: 3 * time.Second,
}, nil)
cleanup = func() {
r.Close()
clean()
}
return r, cleanup, nil
},
}
/*
The test loads a snapshot file to construct the swarm network,
assuming that the snapshot file identifies a healthy
kademlia network. Nevertheless a health check runs in the
simulation's `action` function.
The snapshot should have 'streamer' in its service list.
*/
func runFileRetrievalTest(nodeCount int) error {
sim := simulation.New(retrievalSimServiceMap)
defer sim.Close()
log.Info("Initializing test config")
conf := &synctestConfig{}
//map of discover ID to indexes of chunks expected at that ID
conf.idToChunksMap = make(map[enode.ID][]int)
//map of overlay address to discover ID
conf.addrToIDMap = make(map[string]enode.ID)
//array where the generated chunk hashes will be stored
conf.hashes = make([]storage.Address, 0)
err := sim.UploadSnapshot(fmt.Sprintf("testing/snapshot_%d.json", nodeCount))
if err != nil {
return err
}
ctx, cancelSimRun := context.WithTimeout(context.Background(), 3*time.Minute)
defer cancelSimRun()
result := sim.Run(ctx, func(ctx context.Context, sim *simulation.Simulation) error {
nodeIDs := sim.UpNodeIDs()
for _, n := range nodeIDs {
//get the kademlia overlay address from this ID
a := n.Bytes()
//append it to the array of all overlay addresses
conf.addrs = append(conf.addrs, a)
//the proximity calculation is on overlay addr,
//the p2p/simulations check func triggers on enode.ID,
//so we need to know which overlay addr maps to which nodeID
conf.addrToIDMap[string(a)] = n
}
//an array for the random files
var randomFiles []string
//channel to signal when the upload has finished
//uploadFinished := make(chan struct{})
//channel to trigger new node checks
conf.hashes, randomFiles, err = uploadFilesToNodes(sim)
if err != nil {
return err
}
if _, err := sim.WaitTillHealthy(ctx); err != nil {
return err
}
// File retrieval check is repeated until all uploaded files are retrieved from all nodes
// or until the timeout is reached.
REPEAT:
for {
for _, id := range nodeIDs {
//for each expected file, check if it is in the local store
item, ok := sim.NodeItem(id, bucketKeyFileStore)
if !ok {
return fmt.Errorf("No filestore")
}
fileStore := item.(*storage.FileStore)
//check all chunks
for i, hash := range conf.hashes {
reader, _ := fileStore.Retrieve(context.TODO(), hash)
//check that we can read the file size and that it corresponds to the generated file size
if s, err := reader.Size(ctx, nil); err != nil || s != int64(len(randomFiles[i])) {
log.Debug("Retrieve error", "err", err, "hash", hash, "nodeId", id)
time.Sleep(500 * time.Millisecond)
continue REPEAT
}
log.Debug(fmt.Sprintf("File with root hash %x successfully retrieved", hash))
}
}
return nil
}
})
if result.Error != nil {
return result.Error
}
return nil
}
/*
The test generates the given number of chunks.
The test loads a snapshot file to construct the swarm network,
assuming that the snapshot file identifies a healthy
kademlia network. Nevertheless a health check runs in the
simulation's `action` function.
The snapshot should have 'streamer' in its service list.
*/
func runRetrievalTest(t *testing.T, chunkCount int, nodeCount int) error {
t.Helper()
sim := simulation.New(retrievalSimServiceMap)
defer sim.Close()
conf := &synctestConfig{}
//map of discover ID to indexes of chunks expected at that ID
conf.idToChunksMap = make(map[enode.ID][]int)
//map of overlay address to discover ID
conf.addrToIDMap = make(map[string]enode.ID)
//array where the generated chunk hashes will be stored
conf.hashes = make([]storage.Address, 0)
err := sim.UploadSnapshot(fmt.Sprintf("testing/snapshot_%d.json", nodeCount))
if err != nil {
return err
}
ctx := context.Background()
result := sim.Run(ctx, func(ctx context.Context, sim *simulation.Simulation) error {
nodeIDs := sim.UpNodeIDs()
for _, n := range nodeIDs {
//get the kademlia overlay address from this ID
a := n.Bytes()
//append it to the array of all overlay addresses
conf.addrs = append(conf.addrs, a)
//the proximity calculation is on overlay addr,
//the p2p/simulations check func triggers on enode.ID,
//so we need to know which overlay addr maps to which nodeID
conf.addrToIDMap[string(a)] = n
}
//this is the node selected for upload
node := sim.Net.GetRandomUpNode()
item, ok := sim.NodeItem(node.ID(), bucketKeyStore)
if !ok {
return fmt.Errorf("No localstore")
}
lstore := item.(*storage.LocalStore)
conf.hashes, err = uploadFileToSingleNodeStore(node.ID(), chunkCount, lstore)
if err != nil {
return err
}
if _, err := sim.WaitTillHealthy(ctx); err != nil {
return err
}
// File retrieval check is repeated until all uploaded files are retrieved from all nodes
// or until the timeout is reached.
REPEAT:
for {
for _, id := range nodeIDs {
//for each expected chunk, check if it is in the local store
//check on the node's FileStore (netstore)
item, ok := sim.NodeItem(id, bucketKeyFileStore)
if !ok {
return fmt.Errorf("No filestore")
}
fileStore := item.(*storage.FileStore)
//check all chunks
for _, hash := range conf.hashes {
reader, _ := fileStore.Retrieve(context.TODO(), hash)
//check that we can read the chunk size and that it corresponds to the generated chunk size
if s, err := reader.Size(ctx, nil); err != nil || s != int64(chunkSize) {
log.Debug("Retrieve error", "err", err, "hash", hash, "nodeId", id, "size", s)
time.Sleep(500 * time.Millisecond)
continue REPEAT
}
log.Debug(fmt.Sprintf("Chunk with root hash %x successfully retrieved", hash))
}
}
// all nodes and files found, exit loop and return without error
return nil
}
})
if result.Error != nil {
return result.Error
}
return nil
}