// createMergedSnapshotMessage creates a snapshot message that contains: raft status (term, conf),
// a snapshot of v2 store inside raft.Snapshot as []byte, a snapshot of v3 KV in the top level message
// as ReadCloser.
func (s *EtcdServer) createMergedSnapshotMessage(m raftpb.Message, snapi uint64, confState raftpb.ConfState) snap.Message {
snapt, err := s.r.raftStorage.Term(snapi)
if err != nil {
log.Panicf("get term should never fail: %v", err)
}
// get a snapshot of v2 store as []byte
clone := s.store.Clone()
d, err := clone.SaveNoCopy()
if err != nil {
plog.Panicf("store save should never fail: %v", err)
}
// get a snapshot of v3 KV as readCloser
rc := newSnapshotReaderCloser(s.kv.Snapshot())
// put the []byte snapshot of store into raft snapshot and return the merged snapshot with
// KV readCloser snapshot.
snapshot := raftpb.Snapshot{
Metadata: raftpb.SnapshotMetadata{
Index: snapi,
Term: snapt,
ConfState: confState,
},
Data: d,
}
m.Snapshot = snapshot
return snap.Message{
Message: m,
ReadCloser: rc,
}
}
// sendAppend sends RRPC, with entries to the given peer.
func (r *raft) sendAppend(to uint64) {
pr := r.prs[to]
m := pb.Message{}
m.To = to
if r.needSnapshot(pr.next) {
m.Type = pb.MsgSnap
snapshot, err := r.raftLog.snapshot()
if err != nil {
panic(err) // TODO(bdarnell)
}
if IsEmptySnap(snapshot) {
panic("need non-empty snapshot")
}
m.Snapshot = snapshot
sindex, sterm := snapshot.Metadata.Index, snapshot.Metadata.Term
log.Printf("raft: %x [firstindex: %d, commit: %d] sent snapshot[index: %d, term: %d] to %x [%s]",
r.id, r.raftLog.firstIndex(), r.Commit, sindex, sterm, to, pr)
} else {
m.Type = pb.MsgApp
m.Index = pr.next - 1
m.LogTerm = r.raftLog.term(pr.next - 1)
m.Entries = r.raftLog.entries(pr.next)
m.Commit = r.raftLog.committed
// optimistically increase the next if the follower
// has been matched.
if n := len(m.Entries); pr.match != 0 && n != 0 {
pr.optimisticUpdate(m.Entries[n-1].Index)
}
}
r.send(m)
}
// sendAppend sends RRPC, with entries to the given peer.
func (r *raft) sendAppend(to uint64) {
pr := r.prs[to]
m := pb.Message{}
m.To = to
if r.needSnapshot(pr.next) {
m.Type = pb.MsgSnap
snapshot, err := r.raftLog.snapshot()
if err != nil {
panic(err) // TODO(bdarnell)
}
if IsEmptySnap(snapshot) {
panic("need non-empty snapshot")
}
m.Snapshot = snapshot
} else {
m.Type = pb.MsgApp
m.Index = pr.next - 1
m.LogTerm = r.raftLog.term(pr.next - 1)
m.Entries = r.raftLog.entries(pr.next)
m.Commit = r.raftLog.committed
// optimistically increase the next if the follower
// has been matched.
if n := len(m.Entries); pr.match != 0 && n != 0 {
pr.optimisticUpdate(m.Entries[n-1].Index)
}
}
r.send(m)
}
// createMergedSnapshotMessage creates a snapshot message that contains: raft status (term, conf),
// a snapshot of v2 store inside raft.Snapshot as []byte, a snapshot of v3 KV in the top level message
// as ReadCloser.
func (s *EtcdServer) createMergedSnapshotMessage(m raftpb.Message, snapi uint64, confState raftpb.ConfState) snap.Message {
snapt, err := s.r.raftStorage.Term(snapi)
if err != nil {
log.Panicf("get term should never fail: %v", err)
}
// get a snapshot of v2 store as []byte
clone := s.store.Clone()
d, err := clone.SaveNoCopy()
if err != nil {
plog.Panicf("store save should never fail: %v", err)
}
// commit kv to write metadata(for example: consistent index).
s.KV().Commit()
dbsnap := s.be.Snapshot()
// get a snapshot of v3 KV as readCloser
rc := newSnapshotReaderCloser(dbsnap)
// put the []byte snapshot of store into raft snapshot and return the merged snapshot with
// KV readCloser snapshot.
snapshot := raftpb.Snapshot{
Metadata: raftpb.SnapshotMetadata{
Index: snapi,
Term: snapt,
ConfState: confState,
},
Data: d,
}
m.Snapshot = snapshot
return *snap.NewMessage(m, rc, dbsnap.Size())
}
// sendAppend sends RPC, with entries to the given peer.
func (r *raft) sendAppend(to uint64) {
pr := r.prs[to]
if pr.isPaused() {
return
}
m := pb.Message{}
m.To = to
term, errt := r.raftLog.term(pr.Next - 1)
ents, erre := r.raftLog.entries(pr.Next, r.maxMsgSize)
if errt != nil || erre != nil { // send snapshot if we failed to get term or entries
if !pr.RecentActive {
r.logger.Debugf("ignore sending snapshot to %x since it is not recently active", to)
return
}
m.Type = pb.MsgSnap
snapshot, err := r.raftLog.snapshot()
if err != nil {
if err == ErrSnapshotTemporarilyUnavailable {
r.logger.Debugf("%x failed to send snapshot to %x because snapshot is temporarily unavailable", r.id, to)
return
}
panic(err) // TODO(bdarnell)
}
if IsEmptySnap(snapshot) {
panic("need non-empty snapshot")
}
m.Snapshot = snapshot
sindex, sterm := snapshot.Metadata.Index, snapshot.Metadata.Term
r.logger.Debugf("%x [firstindex: %d, commit: %d] sent snapshot[index: %d, term: %d] to %x [%s]",
r.id, r.raftLog.firstIndex(), r.raftLog.committed, sindex, sterm, to, pr)
pr.becomeSnapshot(sindex)
r.logger.Debugf("%x paused sending replication messages to %x [%s]", r.id, to, pr)
} else {
m.Type = pb.MsgApp
m.Index = pr.Next - 1
m.LogTerm = term
m.Entries = ents
m.Commit = r.raftLog.committed
if n := len(m.Entries); n != 0 {
switch pr.State {
// optimistically increase the next when in ProgressStateReplicate
case ProgressStateReplicate:
last := m.Entries[n-1].Index
pr.optimisticUpdate(last)
pr.ins.add(last)
case ProgressStateProbe:
pr.pause()
default:
r.logger.Panicf("%x is sending append in unhandled state %s", r.id, pr.State)
}
}
}
r.send(m)
}
// sendAppend sends RRPC, with entries to the given peer.
func (r *raft) sendAppend(to int64) {
pr := r.prs[to]
m := pb.Message{}
m.To = to
m.Index = pr.next - 1
if r.needSnapshot(m.Index) {
m.Type = msgSnap
m.Snapshot = r.raftLog.snapshot
} else {
m.Type = msgApp
m.LogTerm = r.raftLog.term(pr.next - 1)
m.Entries = r.raftLog.entries(pr.next)
m.Commit = r.raftLog.committed
}
r.send(m)
}
// sendAppend sends RRPC, with entries to the given peer.
func (r *raft) sendAppend(to uint64) {
pr := r.prs[to]
if pr.isPaused() {
return
}
m := pb.Message{}
m.To = to
if r.needSnapshot(pr.Next) {
m.Type = pb.MsgSnap
snapshot, err := r.raftLog.snapshot()
if err != nil {
panic(err) // TODO(bdarnell)
}
if IsEmptySnap(snapshot) {
panic("need non-empty snapshot")
}
m.Snapshot = snapshot
sindex, sterm := snapshot.Metadata.Index, snapshot.Metadata.Term
raftLogger.Infof("%x [firstindex: %d, commit: %d] sent snapshot[index: %d, term: %d] to %x [%s]",
r.id, r.raftLog.firstIndex(), r.Commit, sindex, sterm, to, pr)
pr.becomeSnapshot(sindex)
raftLogger.Infof("%x paused sending replication messages to %x [%s]", r.id, to, pr)
} else {
m.Type = pb.MsgApp
m.Index = pr.Next - 1
m.LogTerm = r.raftLog.term(pr.Next - 1)
m.Entries = r.raftLog.entries(pr.Next, r.maxMsgSize)
m.Commit = r.raftLog.committed
if n := len(m.Entries); n != 0 {
switch pr.State {
// optimistically increase the next when in ProgressStateReplicate
case ProgressStateReplicate:
last := m.Entries[n-1].Index
pr.optimisticUpdate(last)
pr.ins.add(last)
case ProgressStateProbe:
pr.pause()
default:
raftLogger.Panicf("%x is sending append in unhandled state %s", r.id, pr.State)
}
}
}
r.send(m)
}
// sendAppend sends RRPC, with entries to the given peer.
func (r *raft) sendAppend(to uint64) {
pr := r.prs[to]
if pr.shouldWait() {
log.Printf("raft: %x ignored sending %s to %x [%s]", r.id, pb.MsgApp, to, pr)
return
}
m := pb.Message{}
m.To = to
if r.needSnapshot(pr.next) {
m.Type = pb.MsgSnap
snapshot, err := r.raftLog.snapshot()
if err != nil {
panic(err) // TODO(bdarnell)
}
if IsEmptySnap(snapshot) {
panic("need non-empty snapshot")
}
m.Snapshot = snapshot
sindex, sterm := snapshot.Metadata.Index, snapshot.Metadata.Term
log.Printf("raft: %x [firstindex: %d, commit: %d] sent snapshot[index: %d, term: %d] to %x [%s]",
r.id, r.raftLog.firstIndex(), r.Commit, sindex, sterm, to, pr)
pr.waitSet(r.electionTimeout)
} else {
m.Type = pb.MsgApp
m.Index = pr.next - 1
m.LogTerm = r.raftLog.term(pr.next - 1)
m.Entries = r.raftLog.entries(pr.next)
m.Commit = r.raftLog.committed
// optimistically increase the next if the follower
// has been matched.
if n := len(m.Entries); pr.match != 0 && n != 0 {
pr.optimisticUpdate(m.Entries[n-1].Index)
} else if pr.match == 0 {
// TODO (xiangli): better way to find out if the follower is in good path or not
// a follower might be in bad path even if match != 0, since we optimistically
// increase the next.
pr.waitSet(r.heartbeatTimeout)
}
}
r.send(m)
}
