raft: add conf safety

To make configuration change safe without adding configuration protocol:

1. We only allow to add/remove one node at a time.

2. We only allow one uncommitted configuration entry in the log.

These two rules can make sure there is no disjoint quorums in both current cluster and the
future(after applied any number of committed entries or uncommitted entries in log) clusters.

We add a type field in Entry structure for two reasons:

1. Statemachine needs to know if there is a pending configuration change.

2. Configuration entry should be executed by raft package rather application who is using raft.
This commit is contained in:
Xiang Li 2014-06-05 13:00:18 -07:00 committed by Yicheng Qin
parent 853a458a0d
commit c03fbf68d6
4 changed files with 62 additions and 16 deletions

View File

@ -1,6 +1,12 @@
package raft package raft
const (
normal int = iota
config
)
type Entry struct { type Entry struct {
Type int
Term int Term int
Data []byte Data []byte
} }

View File

@ -33,7 +33,7 @@ func New(addr int, peer []int, heartbeat, election tick) *Node {
// Propose asynchronously proposes data be applied to the underlying state machine. // Propose asynchronously proposes data be applied to the underlying state machine.
func (n *Node) Propose(data []byte) { func (n *Node) Propose(data []byte) {
m := Message{Type: msgProp, Data: data} m := Message{Type: msgProp, Entries: []Entry{{Data: data}}}
n.Step(m) n.Step(m)
} }

View File

@ -63,7 +63,6 @@ type Message struct {
PrevTerm int PrevTerm int
Entries []Entry Entries []Entry
Commit int Commit int
Data []byte
} }
type index struct { type index struct {
@ -103,6 +102,9 @@ type stateMachine struct {
// the leader addr // the leader addr
lead int lead int
// pending reconfiguration
pendingConf bool
} }
func newStateMachine(addr int, peer []int) *stateMachine { func newStateMachine(addr int, peer []int) *stateMachine {
@ -254,9 +256,23 @@ func (sm *stateMachine) Step(m Message) {
sm.bcastAppend() sm.bcastAppend()
return return
case msgProp: case msgProp:
if len(m.Entries) != 1 {
panic("unexpected length(entries) of a msgProp")
}
switch sm.lead { switch sm.lead {
case sm.addr: case sm.addr:
sm.log.append(sm.log.lastIndex(), Entry{Term: sm.term, Data: m.Data}) e := m.Entries[0]
if e.Type == config {
if sm.pendingConf {
// todo: deny
return
}
sm.pendingConf = true
}
e.Term = sm.term
sm.log.append(sm.log.lastIndex(), e)
sm.ins[sm.addr].update(sm.log.lastIndex()) sm.ins[sm.addr].update(sm.log.lastIndex())
sm.maybeCommit() sm.maybeCommit()
sm.bcastAppend() sm.bcastAppend()

View File

@ -45,16 +45,17 @@ func TestLogReplication(t *testing.T) {
{ {
newNetwork(nil, nil, nil), newNetwork(nil, nil, nil),
[]Message{ []Message{
{To: 0, Type: msgProp, Data: []byte("somedata")}, {To: 0, Type: msgProp, Entries: []Entry{{Data: []byte("somedata")}}},
}, },
1, 1,
}, },
{ {
newNetwork(nil, nil, nil), newNetwork(nil, nil, nil),
[]Message{ []Message{
{To: 0, Type: msgProp, Data: []byte("somedata")},
{To: 0, Type: msgProp, Entries: []Entry{{Data: []byte("somedata")}}},
{To: 1, Type: msgHup}, {To: 1, Type: msgHup},
{To: 1, Type: msgProp, Data: []byte("somedata")}, {To: 1, Type: msgProp, Entries: []Entry{{Data: []byte("somedata")}}},
}, },
2, 2,
}, },
@ -82,8 +83,8 @@ func TestLogReplication(t *testing.T) {
} }
} }
for k, m := range props { for k, m := range props {
if !bytes.Equal(ents[k].Data, m.Data) { if !bytes.Equal(ents[k].Data, m.Entries[0].Data) {
t.Errorf("#%d.%d: data = %d, want %d", i, j, ents[k].Data, m.Data) t.Errorf("#%d.%d: data = %d, want %d", i, j, ents[k].Data, m.Entries[0].Data)
} }
} }
} }
@ -93,8 +94,8 @@ func TestLogReplication(t *testing.T) {
func TestSingleNodeCommit(t *testing.T) { func TestSingleNodeCommit(t *testing.T) {
tt := newNetwork(nil) tt := newNetwork(nil)
tt.send(Message{To: 0, Type: msgHup}) tt.send(Message{To: 0, Type: msgHup})
tt.send(Message{To: 0, Type: msgProp, Data: []byte("some data")}) tt.send(Message{To: 0, Type: msgProp, Entries: []Entry{{Data: []byte("some data")}}})
tt.send(Message{To: 0, Type: msgProp, Data: []byte("some data")}) tt.send(Message{To: 0, Type: msgProp, Entries: []Entry{{Data: []byte("some data")}}})
sm := tt.peers[0].(*stateMachine) sm := tt.peers[0].(*stateMachine)
if sm.log.committed != 2 { if sm.log.committed != 2 {
@ -111,8 +112,8 @@ func TestCannotCommitWithoutNewTermEntry(t *testing.T) {
tt.cut(0, 3) tt.cut(0, 3)
tt.cut(0, 4) tt.cut(0, 4)
tt.send(Message{To: 0, Type: msgProp, Data: []byte("some data")}) tt.send(Message{To: 0, Type: msgProp, Entries: []Entry{{Data: []byte("some data")}}})
tt.send(Message{To: 0, Type: msgProp, Data: []byte("some data")}) tt.send(Message{To: 0, Type: msgProp, Entries: []Entry{{Data: []byte("some data")}}})
sm := tt.peers[0].(*stateMachine) sm := tt.peers[0].(*stateMachine)
if sm.log.committed != 0 { if sm.log.committed != 0 {
@ -135,7 +136,7 @@ func TestCannotCommitWithoutNewTermEntry(t *testing.T) {
// after append a entry from the current term, all entries // after append a entry from the current term, all entries
// should be committed // should be committed
tt.send(Message{To: 1, Type: msgProp, Data: []byte("some data")}) tt.send(Message{To: 1, Type: msgProp, Entries: []Entry{{Data: []byte("some data")}}})
if sm.log.committed != 3 { if sm.log.committed != 3 {
t.Errorf("committed = %d, want %d", sm.log.committed, 3) t.Errorf("committed = %d, want %d", sm.log.committed, 3)
} }
@ -197,7 +198,7 @@ func TestCandidateConcede(t *testing.T) {
data := []byte("force follower") data := []byte("force follower")
// send a proposal to 2 to flush out a msgApp to 0 // send a proposal to 2 to flush out a msgApp to 0
tt.send(Message{To: 2, Type: msgProp, Data: data}) tt.send(Message{To: 2, Type: msgProp, Entries: []Entry{{Data: data}}})
a := tt.peers[0].(*stateMachine) a := tt.peers[0].(*stateMachine)
if g := a.state; g != stateFollower { if g := a.state; g != stateFollower {
@ -284,7 +285,7 @@ func TestProposal(t *testing.T) {
// promote 0 the leader // promote 0 the leader
send(Message{To: 0, Type: msgHup}) send(Message{To: 0, Type: msgHup})
send(Message{To: 0, Type: msgProp, Data: data}) send(Message{To: 0, Type: msgProp, Entries: []Entry{{Data: data}}})
wantLog := newLog() wantLog := newLog()
if tt.success { if tt.success {
@ -320,7 +321,7 @@ func TestProposalByProxy(t *testing.T) {
tt.send(Message{To: 0, Type: msgHup}) tt.send(Message{To: 0, Type: msgHup})
// propose via follower // propose via follower
tt.send(Message{To: 1, Type: msgProp, Data: []byte("somedata")}) tt.send(Message{To: 1, Type: msgProp, Entries: []Entry{{Data: []byte("somedata")}}})
wantLog := &log{ents: []Entry{{}, {Term: 1, Data: data}}, committed: 1} wantLog := &log{ents: []Entry{{}, {Term: 1, Data: data}}, committed: 1}
base := ltoa(wantLog) base := ltoa(wantLog)
@ -491,6 +492,29 @@ func TestStateTransition(t *testing.T) {
} }
} }
func TestConf(t *testing.T) {
sm := newStateMachine(0, []int{0})
sm.becomeCandidate()
sm.becomeLeader()
sm.Step(Message{Type: msgProp, Entries: []Entry{{Type: config}}})
if sm.log.lastIndex() != 1 {
t.Errorf("lastindex = %d, want %d", sm.log.lastIndex(), 1)
}
if !sm.pendingConf {
t.Errorf("pendingConf = %v, want %v", sm.pendingConf, true)
}
if sm.log.ents[1].Type != config {
t.Errorf("type = %d, want %d", sm.log.ents[1].Type, config)
}
// deny the second configuration change request if there is a pending one
sm.Step(Message{Type: msgProp, Entries: []Entry{{Type: config}}})
if sm.log.lastIndex() != 1 {
t.Errorf("lastindex = %d, want %d", sm.log.lastIndex(), 1)
}
}
func TestAllServerStepdown(t *testing.T) { func TestAllServerStepdown(t *testing.T) {
tests := []stateType{stateFollower, stateCandidate, stateLeader} tests := []stateType{stateFollower, stateCandidate, stateLeader}