Skip to content

NRG: Parallel catchups can truncate committed #7424

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Merged
neilalexander merged 1 commit into from
Oct 13, 2025
+60 −3
Merged

NRG: Parallel catchups can truncate committed #7424

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
1 commit
Select commit
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
6 changes: 3 additions & 3 deletions server/raft.go
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -3594,9 +3594,9 @@ func (n *raft) processAppendEntry(ae *appendEntry, sub *subscription) {
isNew := sub != nil && sub == n.aesub

// If we are/were catching up ignore old catchup subs, but only if catching up from an older server
// that doesn't send the leader term when catching up. We can reject old catchups from newer subs
// later, just by checking the append entry is on the correct term.
if !isNew && sub != nil && ae.lterm == 0 && (!catchingUp || sub != n.catchup.sub) {
// that doesn't send the leader term when catching up or if we would truncate as a result.
// We can reject old catchups from newer subs later, just by checking the append entry is on the correct term.
if !isNew && sub != nil && (ae.lterm == 0 || ae.pindex < n.pindex) && (!catchingUp || sub != n.catchup.sub) {
n.Unlock()
n.debug("AppendEntry ignoring old entry from previous catchup")
return
Expand Down
57 changes: 57 additions & 0 deletions server/raft_test.go
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -3696,6 +3696,63 @@ func TestNRGLostQuorum(t *testing.T) {
require_False(t, n.lostQuorum())
}

func TestNRGParallelCatchupRollback(t *testing.T) {
n, cleanup := initSingleMemRaftNode(t)
defer cleanup()

// Create a sample entry, the content doesn't matter, just that it's stored.
esm := encodeStreamMsgAllowCompress("foo", "_INBOX.foo", nil, nil, 0, 0, true)
entries := []*Entry{newEntry(EntryNormal, esm)}

nats0 := "S1Nunr6R" // "nats-0"

aeReply := "$TEST"
nc, err := nats.Connect(n.s.ClientURL(), nats.UserInfo("admin", "s3cr3t!"))
require_NoError(t, err)
defer nc.Close()

sub, err := nc.SubscribeSync(aeReply)
require_NoError(t, err)
defer sub.Drain()
require_NoError(t, nc.Flush())

// Timeline
aeMsg1 := encode(t, &appendEntry{leader: nats0, term: 1, lterm: 1, commit: 0, pterm: 0, pindex: 0, entries: entries})
aeMsg2 := encode(t, &appendEntry{leader: nats0, term: 1, lterm: 1, commit: 0, pterm: 1, pindex: 1, entries: entries})
aeHeartbeat := encode(t, &appendEntry{leader: nats0, term: 1, commit: 0, pterm: 1, pindex: 2, entries: nil, reply: aeReply})

// Trigger a catchup.
n.processAppendEntry(aeMsg2, n.aesub)
require_Equal(t, n.pindex, 0)
require_NotNil(t, n.catchup)
require_Equal(t, n.catchup.cterm, aeMsg2.pterm)
require_Equal(t, n.catchup.cindex, aeMsg2.pindex)
csub := n.catchup.sub

// Receive the missed messages.
n.processAppendEntry(aeMsg1, csub)
require_Equal(t, n.pindex, 1)
n.processAppendEntry(aeMsg2, csub)
require_Equal(t, n.pindex, 2)
require_True(t, n.catchup == nil)

// Should respond to the heartbeat and allow the leader to commit.
n.processAppendEntry(aeHeartbeat, n.aesub)
msg, err := sub.NextMsg(time.Second)
require_NoError(t, err)
ar := decodeAppendEntryResponse(msg.Data)
require_NotNil(t, ar)
require_True(t, ar.success)
require_Equal(t, ar.term, aeHeartbeat.term)
require_Equal(t, ar.index, aeHeartbeat.pindex)

// Now replay a message that was already received as a catchup entry.
// Likely due to running multiple catchups in parallel.
// Since our WAL is already ahead, we should not truncate based on this.
n.processAppendEntry(aeMsg1, csub)
require_Equal(t, n.pindex, 2)
}

// This is a RaftChainOfBlocks test where a block is proposed and then we wait for all replicas to apply it before
// proposing the next one.
// The test may fail if:
Expand Down