mirrors/seaweedfs
1
0
Fork 0
mirror of https://github.com/seaweedfs/seaweedfs.git synced 2026-10-21 22:12:36 +00:00
Code Issues Packages Projects Releases Wiki Activity

master: treat rp-only layout differences as an in-flight replication change (#11763)

Browse Source 
volume.configure.replication rolls through nodes one at a time, so a
volume legitimately reports the old and new replica placement while it
settles. Layouts differing only in rp within the same collection are a
supported transition, not a reused id.
This commit is contained in:
Chris Lu
2026-10-20 07:35:41 -07:00
parent d78fccf453
commit 9f1febb2ac
2 changed files with 64 additions and 5 deletions
Download Patch File Download Diff File Expand all files Collapse all files
weed/topology/duplicate_volume_id_test.go
+52
View File
@@ -133,3 +133,55 @@ func TestQuarantineSurvivesResync(t *testing.T) {
t.Fatal("resync must not clear the quarantine")
}
}
func TestReplicationChangeIsNotDuplicate(t *testing.T) {
topo := NewTopology("topo", nil, 128, 3, false)
rpOld, _ := super_block.NewReplicaPlacementFromString("000")
rpNew, _ := super_block.NewReplicaPlacementFromString("001")
dn1 := NewDataNode("server1")
dn1.Ip = "10.0.0.1"
dn1.Port = 8080
dn2 := NewDataNode("server2")
dn2.Ip = "10.0.0.2"
dn2.Port = 8080
vOld := storage.VolumeInfo{Id: 7, Collection: "a", ReplicaPlacement: rpOld, Ttl: needle.EMPTY_TTL}
vNew := storage.VolumeInfo{Id: 7, Collection: "a", ReplicaPlacement: rpNew, Ttl: needle.EMPTY_TTL}
dn1.UpdateVolumes([]storage.VolumeInfo{vOld})
dn2.UpdateVolumes([]storage.VolumeInfo{vNew})
topo.RegisterVolumeLayout(vOld, dn1)
topo.RegisterVolumeLayout(vNew, dn2)
topo.flagDuplicateVolumeId(vNew, dn2)
if topo.isQuarantined(dn2, vNew.Id) || topo.isQuarantined(dn1, vOld.Id) {
t.Fatal("an rp-only layout difference is a replication change in flight, not a duplicate")
}
}
func TestDuplicateVolumeIdAcrossTtls(t *testing.T) {
topo := NewTopology("topo", nil, 128, 3, false)
rp, _ := super_block.NewReplicaPlacementFromString("000")
ttl, _ := needle.ReadTTL("3d")
dn1 := NewDataNode("server1")
dn1.Ip = "10.0.0.1"
dn1.Port = 8080
dn2 := NewDataNode("server2")
dn2.Ip = "10.0.0.2"
dn2.Port = 8080
v1 := storage.VolumeInfo{Id: 7, Collection: "a", ReplicaPlacement: rp, Ttl: needle.EMPTY_TTL}
v2 := storage.VolumeInfo{Id: 7, Collection: "a", ReplicaPlacement: rp, Ttl: ttl}
dn1.UpdateVolumes([]storage.VolumeInfo{v1})
dn2.UpdateVolumes([]storage.VolumeInfo{v2})
topo.RegisterVolumeLayout(v1, dn1)
topo.RegisterVolumeLayout(v2, dn2)
topo.flagDuplicateVolumeId(v2, dn2)
if !topo.isQuarantined(dn2, v2.Id) || !topo.isQuarantined(dn1, v1.Id) {
t.Fatal("same collection with a different ttl is still a reused volume id")
}
}
weed/topology/topology.go
+12 -5
View File
@@ -790,6 +790,13 @@ func volumeLayoutKey(collection string, rp *super_block.ReplicaPlacement, ttl *n
return fmt.Sprintf("collection=%q rp=%s ttl=%s disk=%s", collection, rp, ttl, diskType)
}
func ttlEqual(a, b *needle.TTL) bool {
if a == nil || b == nil {
return a == b
}
return a.ToUint32() == b.ToUint32()
}
// findDuplicateVolumeLayouts reports every layout that already holds v.Id on
// data nodes other than dn. Volume ids are global, so the same id held by two
@@ -802,12 +809,12 @@ func (t *Topology) findDuplicateVolumeLayouts(v storage.VolumeInfo, dn *DataNode
var hits []duplicateHit
for _, item := range t.collectionMap.Items() {
c := item.(*Collection)
var same *VolumeLayout
if c.Name == v.Collection {
same, _ = c.GetVolumeLayout(v.ReplicaPlacement, v.Ttl, diskType)
}
for _, vl := range c.GetAllVolumeLayouts() {
if vl == same {
if c.Name == v.Collection && vl.diskType == diskType && ttlEqual(vl.ttl, v.Ttl) {
// The reporter's own layout, or an rp-only difference:
// a replication change rolls through the nodes one at a
// time, so the same volume legitimately reports the old
// and new rp while it settles.
continue
}
var others []*DataNode