What causes "Unknown group for member: %+v" in dgraph-io/dgraph?

Delete the stale Alpha's postings/writedirectory state or point it at the current cluster so it re-registers with a valid group.

dgraph-io/dgraph · error

Unknown group for member: %+v

Error message

Unknown group for member: %+v

What it means

createProposals checks that each member in the incoming state belongs to a group Zero currently knows about. If dstMember.GroupId has no entry in s.state.Groups, the member is referencing an unknown/stale group and the update is rejected.

Source

Thrown at dgraph/cmd/zero/zero.go:360

			return tab
		}
	}
	return nil
}

func (s *Server) createProposals(dst *pb.Group) ([]*pb.ZeroProposal, error) {
	var res []*pb.ZeroProposal
	if len(dst.Members) > 1 {
		return res, errors.Errorf("Create Proposal: Invalid group: %+v", dst)
	}

	s.RLock()
	defer s.RUnlock()
	// There is only one member. We use for loop because we don't know what the mid is.
	for mid, dstMember := range dst.Members {
		group, has := s.state.Groups[dstMember.GroupId]
		if !has {
			return res, errors.Errorf("Unknown group for member: %+v", dstMember)
		}
		srcMember, has := group.Members[mid]
		if !has {
			return res, errors.Errorf("Unknown member: %+v", dstMember)
		}
		if srcMember.Addr != dstMember.Addr ||
			srcMember.Leader != dstMember.Leader {

			proposal := &pb.ZeroProposal{
				Member: dstMember,
			}
			res = append(res, proposal)
		}
		if !dstMember.Leader {
			// Don't continue to tablets if request is not from the leader.
			return res, nil
		}
		if dst.SnapshotTs > group.SnapshotTs {

View on GitHub (pinned to 759e242be6)

Solutions

Delete the stale Alpha's postings/writedirectory state or point it at the current cluster so it re-registers with a valid group.
Check /state on Zero for current group IDs and compare with the Alpha's reported group.
Restart Zero and ensure its raft state is the authoritative, most recent one (don't mix old p/ and zero directories).
Rebuild the cluster membership cleanly if group IDs diverged (backup, wipe, rejoin).

Example fix

// before
alpha.StartWithStaleGroupDir(oldZeroDir)
// after
alpha.CleanStaleMembershipState() // or start with an empty p dir to rejoin
groups := fetchZeroState("http://zero:6080/state")
if !groupExists(groups, alpha.GroupId) { alpha.ResetGroupId() }
alpha.Start()

Defensive patterns

Strategy: validation

Validate before calling

state, _ := http.Get("http://zero:6080/state")
if !state.Groups[alpha.GroupId].exists {
    alpha.ResetMembership(); alpha.Restart()
}

Type guard

func groupKnown(groups map[uint64]*pb.Group, id uint64) bool {
    _, ok := groups[id]
    return ok
}

Try / catch

_, err := zero.UpdateMembership(ctx, state)
if err != nil && strings.Contains(err.Error(), "Unknown group for member") {
    // stale membership: re-register Alpha
    alpha.DeregisterAndRejoin(ctx)
}

Prevention

Never reuse an old Zero raft/watermark directory with freshly built Alpha data.
Compare Alpha group ids against /state after any restore or rebuild.
Remove orphaned Alpha data dirs when rebuilding a cluster.
Keep zero and p directories from the same cluster generation together.

When it happens

Trigger: An Alpha sends a membership update whose GroupId no longer exists in Zero's state — e.g., after the group was removed or Zero's state was reset while the Alpha still believes it belongs to that group.

Common situations: Zero's raft state wiped/restored from an older snapshot while Alphas keep old group IDs; leftover Alpha data directory reused after a cluster rebuild; group was removed via removeNode and the Alpha hasn't re-registered.

Related errors

AI-assisted analysis of dgraph-io/dgraph@759e242be6 (2026-09-01). Data as JSON: /api/errors/51614ad7722e2140. Report an issue: GitHub.