Files
seaweedfs/weed/cluster/lock_manager/hash_ring_test.go
T
Chris Lu f037fc4dce s3: dial the object lock's primary filer directly (#9626)
* s3: dial the object lock's primary filer directly

The S3 object write lock builds a fresh short-lived lock per write, each
starting at the seed filer. When the seed isn't the key's hash-ring primary
the filer forwards the request to the primary, and in multi-cluster setups
that forward crosses clusters on every write.

Give the lock client a view of the filer lock ring, fed by the master's
LockRingUpdate broadcasts the gateway already receives, so it dials the
primary directly. The view tracks filer membership by version; a stale view
stays correct because the filer still forwards as a fallback.

Also send the initial ring snapshot to S3 clients, not just filers.

* s3: subscribe to lock-ring updates before starting the master loop

The master delivers the initial LockRingUpdate once, on connect. Registering the
callback after KeepConnectedToMaster started left a window where that first
update could arrive before the handler was set and be dropped, delaying the ring
view until the next membership change. Build the lock client and register the
callback in the masters block before launching the loop; the filers block reuses
that client (or creates a plain one when no masters are configured).

* lock_manager: build the hash ring in a deterministic server order

rebuildRing ranged over the server set (a map), whose iteration order is
randomized per process. On a vnode hash collision the last writer into
vnodeToServer wins, so two nodes holding the same server set could resolve the
collision to different servers and disagree on the primary for keys near that
slot. Now that the S3 gateway also computes PrimaryForKey, such a disagreement
would route the same key to different filers and defeat per-path serialization.

Iterate the servers in sorted order so the ring is identical on every node with
the same set, regardless of discovery order.

* lock_manager: skip redundant ring rebuilds, trim comments

SetRing now ignores a non-zero version at or below the current one once a ring
exists, so repeated LockRingUpdate broadcasts on reconnect no longer rebuild the
ring.

* s3: hold the lock-ring client on the server for route-by-key

Store the object-write lock client on S3ApiServer so handlers can resolve a
key's owner filer via PrimaryForKey.
2026-05-24 00:40:43 -07:00

209 lines
6.0 KiB
Go

package lock_manager
import (
"fmt"
"math"
"testing"
"github.com/seaweedfs/seaweedfs/weed/pb"
"github.com/stretchr/testify/assert"
)
func TestHashRing_BasicOperations(t *testing.T) {
hr := NewHashRing(50)
// Empty ring
p, b := hr.GetPrimaryAndBackup("key1")
assert.Equal(t, pb.ServerAddress(""), p)
assert.Equal(t, pb.ServerAddress(""), b)
// Single server — no backup
hr.AddServer("filer1:8888")
p, b = hr.GetPrimaryAndBackup("key1")
assert.Equal(t, pb.ServerAddress("filer1:8888"), p)
assert.Equal(t, pb.ServerAddress(""), b)
// Two servers — backup is the other server
hr.AddServer("filer2:8888")
p, b = hr.GetPrimaryAndBackup("key1")
assert.NotEqual(t, p, b)
assert.NotEmpty(t, b)
// Three servers
hr.AddServer("filer3:8888")
p, b = hr.GetPrimaryAndBackup("key1")
assert.NotEqual(t, p, b)
assert.NotEmpty(t, b)
// Remove server
hr.RemoveServer("filer2:8888")
assert.Equal(t, 2, hr.ServerCount())
}
func TestHashRing_DuplicateAddRemove(t *testing.T) {
hr := NewHashRing(50)
hr.AddServer("filer1:8888")
hr.AddServer("filer1:8888") // duplicate
assert.Equal(t, 1, hr.ServerCount())
hr.RemoveServer("filer1:8888")
assert.Equal(t, 0, hr.ServerCount())
hr.RemoveServer("filer1:8888") // remove non-existent
assert.Equal(t, 0, hr.ServerCount())
}
func TestHashRing_SetServers(t *testing.T) {
hr := NewHashRing(50)
hr.SetServers([]pb.ServerAddress{"a:1", "b:2", "c:3"})
assert.Equal(t, 3, hr.ServerCount())
servers := hr.GetServers()
assert.Equal(t, 3, len(servers))
// SetServers replaces
hr.SetServers([]pb.ServerAddress{"x:1", "y:2"})
assert.Equal(t, 2, hr.ServerCount())
}
func TestHashRing_ConsistencyOnRemoval(t *testing.T) {
// The key property of consistent hashing: when a server is removed,
// only keys that mapped to the removed server change.
hr := NewHashRing(50)
servers := []pb.ServerAddress{"filer1:8888", "filer2:8888", "filer3:8888"}
hr.SetServers(servers)
numKeys := 1000
// Record where each key maps before removal
before := make(map[string]pb.ServerAddress, numKeys)
for i := 0; i < numKeys; i++ {
key := fmt.Sprintf("lock-key-%d", i)
before[key] = hr.GetPrimary(key)
}
// Remove filer2
hr.RemoveServer("filer2:8888")
moved := 0
for i := 0; i < numKeys; i++ {
key := fmt.Sprintf("lock-key-%d", i)
after := hr.GetPrimary(key)
if before[key] != after {
// Only keys from filer2 should move
assert.Equal(t, pb.ServerAddress("filer2:8888"), before[key],
"key %s moved from %s to %s, but it wasn't on the removed server", key, before[key], after)
moved++
}
}
// Roughly 1/3 of keys should move (those that were on filer2)
t.Logf("Keys that moved: %d / %d", moved, numKeys)
assert.Greater(t, moved, 0, "some keys should have moved")
assert.Less(t, moved, numKeys, "not all keys should move")
}
func TestHashRing_BackupIsSuccessor(t *testing.T) {
// After removing primary, the backup should become the new primary
hr := NewHashRing(50)
servers := []pb.ServerAddress{"filer1:8888", "filer2:8888", "filer3:8888"}
hr.SetServers(servers)
// For each key, verify that removing the primary makes the backup the new primary
promoted := 0
total := 500
for i := 0; i < total; i++ {
key := fmt.Sprintf("test-lock-%d", i)
primary, backup := hr.GetPrimaryAndBackup(key)
assert.NotEqual(t, primary, backup)
// Temporarily remove primary
hr.RemoveServer(primary)
newPrimary := hr.GetPrimary(key)
if newPrimary == backup {
promoted++
}
// Restore
hr.AddServer(primary)
}
// The backup should become new primary for all keys
assert.Equal(t, total, promoted,
"backup should become new primary for all keys when primary is removed")
}
func TestHashRing_Distribution(t *testing.T) {
hr := NewHashRing(50)
servers := []pb.ServerAddress{"filer1:8888", "filer2:8888", "filer3:8888"}
hr.SetServers(servers)
counts := make(map[pb.ServerAddress]int)
numKeys := 3000
for i := 0; i < numKeys; i++ {
key := fmt.Sprintf("dist-key-%d", i)
p := hr.GetPrimary(key)
counts[p]++
}
expected := float64(numKeys) / float64(len(servers))
for server, count := range counts {
deviation := math.Abs(float64(count)-expected) / expected
t.Logf("Server %s: %d keys (%.1f%% deviation)", server, count, deviation*100)
// Allow up to 40% deviation with 50 vnodes and 3 servers
assert.Less(t, deviation, 0.40,
"server %s has too many or too few keys: %d (expected ~%d)", server, count, int(expected))
}
}
func TestHashRing_GetPrimary(t *testing.T) {
hr := NewHashRing(50)
// Empty ring
assert.Equal(t, pb.ServerAddress(""), hr.GetPrimary("key"))
hr.SetServers([]pb.ServerAddress{"a:1", "b:2"})
// Deterministic: same key always maps to same server
p1 := hr.GetPrimary("mykey")
p2 := hr.GetPrimary("mykey")
assert.Equal(t, p1, p2)
// GetPrimary matches the primary from GetPrimaryAndBackup
primary, _ := hr.GetPrimaryAndBackup("mykey")
assert.Equal(t, primary, hr.GetPrimary("mykey"))
}
// The ring must be identical on every node holding the same server set,
// regardless of the order servers were added or supplied. Build rings several
// ways and assert they agree on the primary for a wide range of keys.
func TestHashRing_OrderIndependent(t *testing.T) {
servers := []pb.ServerAddress{
"filer-a:8888", "filer-b:8888", "filer-c:8888", "filer-d:8888", "filer-e:8888",
}
bySet := NewHashRing(50)
bySet.SetServers(servers)
byReverse := NewHashRing(50)
rev := append([]pb.ServerAddress(nil), servers...)
for i, j := 0, len(rev)-1; i < j; i, j = i+1, j-1 {
rev[i], rev[j] = rev[j], rev[i]
}
byReverse.SetServers(rev)
byAdd := NewHashRing(50)
for _, s := range []pb.ServerAddress{"filer-c:8888", "filer-e:8888", "filer-a:8888", "filer-d:8888", "filer-b:8888"} {
byAdd.AddServer(s)
}
for i := 0; i < 5000; i++ {
key := fmt.Sprintf("s3.object.write:/buckets/b/obj-%d", i)
p := bySet.GetPrimary(key)
if got := byReverse.GetPrimary(key); got != p {
t.Fatalf("reverse-order ring disagrees on %q: %s vs %s", key, got, p)
}
if got := byAdd.GetPrimary(key); got != p {
t.Fatalf("add-order ring disagrees on %q: %s vs %s", key, got, p)
}
}
}