tests: Watch events during linearizability test and compare history

Signed-off-by: Marek Siarkowicz <siarkowicz@google.com>
dependabot/go_modules/go.uber.org/atomic-1.10.0
Marek Siarkowicz 2022-12-24 20:25:17 +01:00
parent 2965817150
commit 6821e226dd
2 changed files with 136 additions and 13 deletions

View File

@ -16,7 +16,6 @@ package linearizability
import (
"context"
"fmt"
"os"
"path/filepath"
"strings"
@ -25,6 +24,7 @@ import (
"time"
"github.com/anishathalye/porcupine"
"github.com/google/go-cmp/cmp"
"golang.org/x/time/rate"
"go.etcd.io/etcd/tests/v3/framework/e2e"
@ -107,29 +107,38 @@ func testLinearizability(ctx context.Context, t *testing.T, config e2e.EtcdProce
t.Fatal(err)
}
defer clus.Close()
ctx, cancel := context.WithCancel(ctx)
trafficCtx, trafficCancel := context.WithCancel(ctx)
go func() {
defer cancel()
err := triggerFailpoints(ctx, t, clus, failpoint)
if err != nil {
t.Error(err)
}
defer trafficCancel()
triggerFailpoints(ctx, t, clus, failpoint)
// Wait second to collect traffic after triggering last failpoint.
time.Sleep(time.Second)
}()
operations := simulateTraffic(ctx, t, clus, traffic)
watchCtx, watchCancel := context.WithCancel(ctx)
var operations []porcupine.Operation
go func() {
defer watchCancel()
operations = simulateTraffic(trafficCtx, t, clus, traffic)
// Wait second to collect watch events after all traffic was sent.
time.Sleep(time.Second)
}()
events := collectClusterWatchEvents(watchCtx, t, clus)
err = clus.Stop()
if err != nil {
t.Error(err)
}
validateEventsMatch(t, events)
checkOperationsAndPersistResults(t, operations, clus)
}
func triggerFailpoints(ctx context.Context, t *testing.T, clus *e2e.EtcdProcessCluster, config FailpointConfig) error {
func triggerFailpoints(ctx context.Context, t *testing.T, clus *e2e.EtcdProcessCluster, config FailpointConfig) {
var err error
successes := 0
failures := 0
for _, proc := range clus.Procs {
if !config.failpoint.Available(proc) {
return fmt.Errorf("failpoint %q not available on %s", config.failpoint.Name(), proc.Config().Name)
t.Errorf("Failpoint %q not available on %s", config.failpoint.Name(), proc.Config().Name)
return
}
}
for successes < config.count && failures < config.retries {
@ -143,10 +152,8 @@ func triggerFailpoints(ctx context.Context, t *testing.T, clus *e2e.EtcdProcessC
successes++
}
if successes < config.count || failures >= config.retries {
return fmt.Errorf("failed to trigger failpoints enough times, err: %v", err)
t.Errorf("failed to trigger failpoints enough times, err: %v", err)
}
time.Sleep(config.waitBetweenTriggers)
return nil
}
type FailpointConfig struct {
@ -203,6 +210,25 @@ type trafficConfig struct {
traffic Traffic
}
func validateEventsMatch(t *testing.T, ops [][]watchEvent) {
// Move longest history to ops[0]
maxLength := len(ops[0])
for i := 1; i < len(ops); i++ {
if len(ops[i]) > maxLength {
maxLength = len(ops[i])
ops[0], ops[i] = ops[i], ops[0]
}
}
for i := 1; i < len(ops); i++ {
length := len(ops[i])
// We compare prefix of watch events, as we are not guaranteed to collect all events from each node.
if diff := cmp.Diff(ops[0][:length], ops[i][:length]); diff != "" {
t.Errorf("Events in watches do not match, %s", diff)
}
}
}
func checkOperationsAndPersistResults(t *testing.T, operations []porcupine.Operation, clus *e2e.EtcdProcessCluster) {
path, err := testResultsDirectory(t)
if err != nil {

View File

@ -0,0 +1,97 @@
// Copyright 2022 The etcd Authors
//
// Licensed under the Apache License, Version 2.0 (the "License");
// you may not use this file except in compliance with the License.
// You may obtain a copy of the License at
//
// http://www.apache.org/licenses/LICENSE-2.0
//
// Unless required by applicable law or agreed to in writing, software
// distributed under the License is distributed on an "AS IS" BASIS,
// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
// See the License for the specific language governing permissions and
// limitations under the License.
package linearizability
import (
"context"
"sync"
"testing"
"time"
"go.uber.org/zap"
"go.etcd.io/etcd/api/v3/mvccpb"
clientv3 "go.etcd.io/etcd/client/v3"
"go.etcd.io/etcd/tests/v3/framework/e2e"
)
func collectClusterWatchEvents(ctx context.Context, t *testing.T, clus *e2e.EtcdProcessCluster) [][]watchEvent {
mux := sync.Mutex{}
var wg sync.WaitGroup
memberEvents := make([][]watchEvent, len(clus.Procs))
for i, member := range clus.Procs {
c, err := clientv3.New(clientv3.Config{
Endpoints: member.EndpointsV3(),
Logger: zap.NewNop(),
DialKeepAliveTime: 1 * time.Millisecond,
DialKeepAliveTimeout: 5 * time.Millisecond,
})
if err != nil {
t.Fatal(err)
}
wg.Add(1)
go func(i int, c *clientv3.Client) {
defer wg.Done()
defer c.Close()
events := collectMemberWatchEvents(ctx, t, c)
mux.Lock()
memberEvents[i] = events
mux.Unlock()
}(i, c)
}
wg.Wait()
return memberEvents
}
func collectMemberWatchEvents(ctx context.Context, t *testing.T, c *clientv3.Client) []watchEvent {
events := []watchEvent{}
var lastRevision int64 = 1
for {
select {
case <-ctx.Done():
return events
default:
}
for resp := range c.Watch(ctx, "", clientv3.WithPrefix(), clientv3.WithRev(lastRevision)) {
lastRevision = resp.Header.Revision
for _, event := range resp.Events {
var op OperationType
switch event.Type {
case mvccpb.PUT:
op = Put
case mvccpb.DELETE:
op = Delete
}
events = append(events, watchEvent{
Op: op,
Key: string(event.Kv.Key),
Value: string(event.Kv.Value),
Revision: event.Kv.ModRevision,
})
}
if resp.Err() != nil {
t.Logf("Watch error: %v", resp.Err())
}
}
}
}
type watchEvent struct {
Op OperationType
Key string
Value string
Revision int64
}