DEPS: upgrade Raft to 72d62a1

[erikgrinaker]

Makes Raft leaders step down when removing themselves from the range.

72d62a1 2023-06-13: step down when leader removes itself [Erik Grinaker]

Epic: none
Release note: None

[cockroach-teamcity]

This change is 

[erikgrinaker]

For some reason, etcd-io/raft#77 breaks leader removals. Should've known better than to make opportunistic cleanups, I suppose. Not sure why it breaks yet, will dig into it tomorrow.

[erikgrinaker]

I think the problem is that the leader steps down before we complete the joint quorum change:

I230627 07:36:53.597613 796 go.etcd.io/raft/v3/raft.go:1886  [T1,n1,s1,r64/1:/{Table/Max-Max}] 2883  1 switched to configuration voters=(2 3) learners=(1)
I230627 07:36:53.597678 796 go.etcd.io/raft/v3/raft.go:835  [T1,n1,s1,r64/1:/{Table/Max-Max}] 2884  1 became follower at term 6
I230627 07:36:53.597749 796 kv/kvserver/replica_raft.go:842  [T1,n1,s1,r64/1:/{Table/Max-Max},raft] 2885  raft leader changed: 1 -> 0
I230627 07:36:53.597739 2901 go.etcd.io/raft/v3/raft.go:1886  [T1,n2,s2,r64/2:/{Table/Max-Max}] 2886  2 switched to configuration voters=(2 3) learners=(1)
I230627 07:36:53.597811 3809 go.etcd.io/raft/v3/raft.go:1886  [T1,n3,s3,r64/3:/{Table/Max-Max}] 2887  3 switched to configuration voters=(2 3) learners=(1)

The leader is now a learner, but it needs to propose and commit the conf change that actually removes it from the range, otherwise noone else will campaign for leadership.

[erikgrinaker]

The problem is a mismatch between Raft and CRDB. In Raft, the leader steps down once it become a learner in a joint config. In CRDB, we don't elect a new leader until the old leader is fully removed from the range -- so when the leader becomes a learner and steps down, the range remains leaderless. This patch fixes it:

--- a/pkg/kv/kvserver/client_raft_test.go
+++ b/pkg/kv/kvserver/client_raft_test.go
diff --git a/pkg/kv/kvserver/replica_raft.go b/pkg/kv/kvserver/replica_raft.go
index d33eece0989..48b1aed4727 100644
--- a/pkg/kv/kvserver/replica_raft.go
+++ b/pkg/kv/kvserver/replica_raft.go
@@ -2570,7 +2570,7 @@ func shouldCampaignAfterConfChange(
                // don't expect to hit this, but let's be defensive.
                return false
        }
-       if _, ok := desc.GetReplicaDescriptorByID(roachpb.ReplicaID(raftStatus.Lead)); ok {
+       if r, ok := desc.GetReplicaDescriptorByID(roachpb.ReplicaID(raftStatus.Lead)); ok && r.IsAnyVoter() {
                // The leader is still in the descriptor.
                return false
        }

Unfortunately, this will break backwards compatibility with 23.1 nodes. We have a few options:

• Roll back the step-down change in etcd/raft.
• Make etcd/raft step-down opt-out or opt-in.
• Modify etcd/raft to step down on removal, not when becoming learner.

The Raft thesis describes leader removal in section 4.2.2, we should probably just do what's specified there, with an option to enable/disable it, for backwards compatibility and migrations.

[erikgrinaker]

Making it opt-in in etcd-io/raft#79

[erikgrinaker]

Closing in favor of #107203. Raft has already been bumped.