add mix version for snapshot e2e case

Signed-off-by: WangXiaoxiao <1141195807@qq.com>
This commit is contained in:
WangXiaoxiao 2022-12-04 22:42:44 +08:00 committed by Benjamin Wang
parent a7e94c4843
commit 8e56e2f5ce
2 changed files with 112 additions and 18 deletions

View File

@ -21,15 +21,15 @@ import (
"time" "time"
"github.com/stretchr/testify/assert" "github.com/stretchr/testify/assert"
"github.com/stretchr/testify/require"
"go.etcd.io/etcd/client/pkg/v3/fileutil" "go.etcd.io/etcd/client/pkg/v3/fileutil"
"go.etcd.io/etcd/tests/v3/framework/config" "go.etcd.io/etcd/tests/v3/framework/config"
"go.etcd.io/etcd/tests/v3/framework/e2e" "go.etcd.io/etcd/tests/v3/framework/e2e"
) )
// TestMixVersionsSendSnapshot tests the mix version send snapshots // TestMixVersionsSnapshotByAddingMember tests the mix version send snapshots by adding member
// TODO(ahrtr): add network partition scenario to trigger snapshots. func TestMixVersionsSnapshotByAddingMember(t *testing.T) {
func TestMixVersionsSendSnapshot(t *testing.T) {
cases := []struct { cases := []struct {
name string name string
clusterVersion e2e.ClusterVersion clusterVersion e2e.ClusterVersion
@ -54,12 +54,12 @@ func TestMixVersionsSendSnapshot(t *testing.T) {
for _, tc := range cases { for _, tc := range cases {
t.Run(tc.name, func(t *testing.T) { t.Run(tc.name, func(t *testing.T) {
mixVersionsSnapshotTest(t, tc.clusterVersion, tc.newInstanceVersion) mixVersionsSnapshotTestByAddingMember(t, tc.clusterVersion, tc.newInstanceVersion)
}) })
} }
} }
func mixVersionsSnapshotTest(t *testing.T, clusterVersion, newInstanceVersion e2e.ClusterVersion) { func mixVersionsSnapshotTestByAddingMember(t *testing.T, clusterVersion, newInstanceVersion e2e.ClusterVersion) {
e2e.BeforeTest(t) e2e.BeforeTest(t)
if !fileutil.Exist(e2e.BinPath.EtcdLastRelease) { if !fileutil.Exist(e2e.BinPath.EtcdLastRelease) {
@ -72,13 +72,10 @@ func mixVersionsSnapshotTest(t *testing.T, clusterVersion, newInstanceVersion e2
e2e.WithSnapshotCount(10), e2e.WithSnapshotCount(10),
e2e.WithVersion(clusterVersion), e2e.WithVersion(clusterVersion),
) )
if err != nil { require.NoError(t, err, "failed to start etcd cluster: %v", err)
t.Fatalf("failed to start etcd cluster: %v", err)
}
defer func() { defer func() {
if err := epc.Close(); err != nil { err := epc.Close()
t.Fatalf("failed to close etcd cluster: %v", err) require.NoError(t, err, "failed to close etcd cluster: %v", err)
}
}() }()
// Write more than SnapshotCount entries to trigger at least a snapshot. // Write more than SnapshotCount entries to trigger at least a snapshot.
@ -86,18 +83,17 @@ func mixVersionsSnapshotTest(t *testing.T, clusterVersion, newInstanceVersion e2
for i := 0; i < 20; i++ { for i := 0; i < 20; i++ {
key := fmt.Sprintf("key-%d", i) key := fmt.Sprintf("key-%d", i)
value := fmt.Sprintf("value-%d", i) value := fmt.Sprintf("value-%d", i)
if err := epc.Client().Put(context.TODO(), key, value, config.PutOptions{}); err != nil { err := epc.Client().Put(context.TODO(), key, value, config.PutOptions{})
t.Fatalf("failed to put %q, error: %v", key, err) require.NoError(t, err, "failed to put %q, error: %v", key, err)
}
} }
// start a new etcd instance, which will receive a snapshot from the leader. // start a new etcd instance, which will receive a snapshot from the leader.
newCfg := *epc.Cfg newCfg := *epc.Cfg
newCfg.Version = newInstanceVersion newCfg.Version = newInstanceVersion
newCfg.SnapshotCatchUpEntries = 10
t.Log("Starting a new etcd instance") t.Log("Starting a new etcd instance")
if err := epc.StartNewProc(context.TODO(), &newCfg, t); err != nil { err = epc.StartNewProc(context.TODO(), &newCfg, t)
t.Fatalf("failed to start the new etcd instance: %v", err) require.NoError(t, err, "failed to start the new etcd instance: %v", err)
}
defer epc.CloseProc(context.TODO(), nil) defer epc.CloseProc(context.TODO(), nil)
// verify all nodes have exact same revision and hash // verify all nodes have exact same revision and hash
@ -123,3 +119,99 @@ func mixVersionsSnapshotTest(t *testing.T, clusterVersion, newInstanceVersion e2
return true return true
}, 10*time.Second, 500*time.Millisecond) }, 10*time.Second, 500*time.Millisecond)
} }
func TestMixVersionsSnapshotByMockingPartition(t *testing.T) {
cases := []struct {
name string
clusterVersion e2e.ClusterVersion
mockPartitionNodeIndex int
}{
{
name: "etcd instance with last version receives snapshot from the leader with current version",
clusterVersion: e2e.MinorityLastVersion,
mockPartitionNodeIndex: 2,
},
}
for _, tc := range cases {
t.Run(tc.name, func(t *testing.T) {
mixVersionsSnapshotTestByMockPartition(t, tc.clusterVersion, tc.mockPartitionNodeIndex)
})
}
}
func mixVersionsSnapshotTestByMockPartition(t *testing.T, clusterVersion e2e.ClusterVersion, mockPartitionNodeIndex int) {
e2e.BeforeTest(t)
if !fileutil.Exist(e2e.BinPath.EtcdLastRelease) {
t.Skipf("%q does not exist", e2e.BinPath.EtcdLastRelease)
}
// Create an etcd cluster with 3 member of MinorityLastVersion
epc, err := e2e.NewEtcdProcessCluster(context.TODO(), t,
e2e.WithClusterSize(3),
e2e.WithSnapshotCount(10),
e2e.WithVersion(clusterVersion),
e2e.WithSnapshotCatchUpEntries(10),
)
require.NoError(t, err, "failed to start etcd cluster: %v", err)
defer func() {
err := epc.Close()
require.NoError(t, err, "failed to close etcd cluster: %v", err)
}()
toPartitionedMember := epc.Procs[mockPartitionNodeIndex]
// Stop and restart the partitioned member
err = toPartitionedMember.Stop()
require.NoError(t, err)
// Write more than SnapshotCount entries to trigger at least a snapshot.
t.Log("Writing 20 keys to the cluster")
for i := 0; i < 20; i++ {
key := fmt.Sprintf("key-%d", i)
value := fmt.Sprintf("value-%d", i)
err := epc.Client().Put(context.TODO(), key, value, config.PutOptions{})
require.NoError(t, err, "failed to put %q, error: %v", key, err)
}
t.Log("Verify logs to check leader has saved snapshot")
leaderEPC := epc.Procs[epc.WaitLeader(t)]
e2e.AssertProcessLogs(t, leaderEPC, "saved snapshot")
// Restart the partitioned member
err = toPartitionedMember.Restart(context.TODO())
require.NoError(t, err)
// verify all nodes have exact same revision and hash
t.Log("Verify all nodes have exact same revision and hash")
assert.Eventually(t, func() bool {
hashKvs, err := epc.Client().HashKV(context.TODO(), 0)
if err != nil {
t.Logf("failed to get HashKV: %v", err)
return false
}
if len(hashKvs) != 3 {
t.Logf("expected 3 hashkv responses, but got: %d", len(hashKvs))
return false
}
if hashKvs[0].Header.Revision != hashKvs[1].Header.Revision {
t.Logf("Got different revisions, [%d, %d]", hashKvs[0].Header.Revision, hashKvs[1].Header.Revision)
return false
}
if hashKvs[1].Header.Revision != hashKvs[2].Header.Revision {
t.Logf("Got different revisions, [%d, %d]", hashKvs[1].Header.Revision, hashKvs[2].Header.Revision)
return false
}
assert.Equal(t, hashKvs[0].Hash, hashKvs[1].Hash)
assert.Equal(t, hashKvs[1].Hash, hashKvs[2].Hash)
return true
}, 10*time.Second, 500*time.Millisecond)
// assert process logs to check snapshot be sent
t.Log("Verify logs to check snapshot be sent from leader to follower")
leaderEPC = epc.Procs[epc.WaitLeader(t)]
e2e.AssertProcessLogs(t, leaderEPC, "sent database snapshot")
}

View File

@ -582,8 +582,10 @@ func (cfg *EtcdProcessClusterConfig) EtcdServerProcessConfig(tb testing.TB, i in
args = append(args, "--experimental-watch-progress-notify-interval", cfg.WatchProcessNotifyInterval.String()) args = append(args, "--experimental-watch-progress-notify-interval", cfg.WatchProcessNotifyInterval.String())
} }
if cfg.SnapshotCatchUpEntries > 0 { if cfg.SnapshotCatchUpEntries > 0 {
if cfg.Version == CurrentVersion || (cfg.Version == MinorityLastVersion && i <= cfg.ClusterSize/2) || (cfg.Version == QuorumLastVersion && i > cfg.ClusterSize/2) {
args = append(args, "--experimental-snapshot-catchup-entries", fmt.Sprintf("%d", cfg.SnapshotCatchUpEntries)) args = append(args, "--experimental-snapshot-catchup-entries", fmt.Sprintf("%d", cfg.SnapshotCatchUpEntries))
} }
}
envVars := map[string]string{} envVars := map[string]string{}
for key, value := range cfg.EnvVars { for key, value := range cfg.EnvVars {
envVars[key] = value envVars[key] = value