Files
seaweedfs/weed/server/filer_pipelined_sender_test.go
T
ed7324611c filer: metrics for metadata subscription delivery (lag, holds, ring fall-offs, claim and watermark ages) (#12773)
* filer: metrics for metadata subscription delivery (lag, holds, ring fall-offs, claim and watermark ages)

* filer: measure subscribe delivery lag after the real send using the mutation stamp, and record a watermark hold once per hold

* filer: skip empty progress markers when sampling subscribe delivery lag

* filer: assert subscribe delivery lag is sampled only after the stream send succeeds

* grafana: panels for the metadata subscription delivery metrics

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>

* filer: record a watermark hold when the read passes the held entry, not when the next hold starts

* filer: use a monotonic clock for the subscribe hold duration

---------

Co-authored-by: Chris Lu <chrislusf@users.noreply.github.com>
Co-authored-by: Claude Opus 5.5 <noreply@anthropic.com>
2026-11-19 12:43:34 -08:00

285 lines
9.5 KiB
Go

package weed_server
import (
"context"
"sync/atomic"
"fmt"
"sync"
"testing"
"time"
"github.com/prometheus/client_golang/prometheus"
dto "github.com/prometheus/client_model/go"
"github.com/seaweedfs/seaweedfs/weed/pb/filer_pb"
"github.com/seaweedfs/seaweedfs/weed/stats"
)
type recordingStream struct {
mu sync.Mutex
slow time.Duration
msgs []*filer_pb.SubscribeMetadataResponse
}
func (s *recordingStream) Send(m *filer_pb.SubscribeMetadataResponse) error {
if s.slow > 0 {
time.Sleep(s.slow)
}
s.mu.Lock()
defer s.mu.Unlock()
// Clone-ish: the sender clears Events after sending, so keep our own view.
copied := &filer_pb.SubscribeMetadataResponse{
TsNs: m.TsNs,
Directory: m.Directory,
EventNotification: m.EventNotification,
LogFileRefs: m.LogFileRefs,
Events: append([]*filer_pb.SubscribeMetadataResponse(nil), m.Events...),
}
s.msgs = append(s.msgs, copied)
return nil
}
func (s *recordingStream) snapshot() []*filer_pb.SubscribeMetadataResponse {
s.mu.Lock()
defer s.mu.Unlock()
return append([]*filer_pb.SubscribeMetadataResponse(nil), s.msgs...)
}
// TestPipelinedSenderRefsNeverBatched pins the wire rules the client depends
// on: a refs message must arrive solo - the client recognizes refs by the
// top-level field and skips the rest of the response, so a refs envelope would
// drop its Events tail, and refs inside Events would be applied as an empty
// event. Everything must arrive, in order, whatever the batcher does.
func TestPipelinedSenderRefsNeverBatched(t *testing.T) {
stream := &recordingStream{slow: 2 * time.Millisecond} // let the queue back up so batching engages
sender := newPipelinedSender(context.Background(), stream, 64, true, "local")
oldTs := time.Now().Add(-time.Hour).UnixNano() // far behind: the batch heuristic fires
var wantOrder []string
send := func(kind string, m *filer_pb.SubscribeMetadataResponse) {
wantOrder = append(wantOrder, kind)
if err := sender.Send(m); err != nil {
t.Fatalf("send %s: %v", kind, err)
}
}
event := func(i int) *filer_pb.SubscribeMetadataResponse {
return &filer_pb.SubscribeMetadataResponse{
TsNs: oldTs + int64(i),
EventNotification: &filer_pb.EventNotification{NewEntry: &filer_pb.Entry{Name: fmt.Sprintf("e%d", i)}},
}
}
refs := func() *filer_pb.SubscribeMetadataResponse {
return &filer_pb.SubscribeMetadataResponse{LogFileRefs: []*filer_pb.LogFileChunkRef{{FilerId: "a"}}}
}
// Interleave backlog events with refs so refs land both between batches
// and mid-drain.
for i := 0; i < 30; i++ {
send("event", event(i))
if i%7 == 3 {
send("refs", refs())
}
}
if err := sender.Close(); err != nil {
t.Fatalf("close: %v", err)
}
var gotOrder []string
for _, m := range stream.snapshot() {
if len(m.LogFileRefs) > 0 {
if len(m.Events) > 0 {
t.Fatal("a refs envelope carried an Events tail; the client drops that tail")
}
if m.EventNotification != nil {
t.Fatal("a refs message doubled as an event envelope")
}
gotOrder = append(gotOrder, "refs")
continue
}
gotOrder = append(gotOrder, "event")
for _, e := range m.Events {
if len(e.LogFileRefs) > 0 {
t.Fatal("refs packed inside Events; the client applies that as an empty event")
}
gotOrder = append(gotOrder, "event")
}
}
if fmt.Sprint(gotOrder) != fmt.Sprint(wantOrder) {
t.Fatalf("delivery order/count changed:\n got %v\nwant %v", gotOrder, wantOrder)
}
}
type blockedStream struct {
release chan struct{}
entered chan struct{}
sends atomic.Int32
}
func (s *blockedStream) Send(*filer_pb.SubscribeMetadataResponse) error {
s.sends.Add(1)
select {
case s.entered <- struct{}{}:
default:
}
<-s.release
return nil
}
// A subscriber that stops reading wedges stream.Send. The subscription
// context must release Send and Close, so the handler can return and end the
// stream during filer shutdown.
func TestPipelinedSenderUnblocksOnCancel(t *testing.T) {
stream := &blockedStream{release: make(chan struct{})}
defer close(stream.release)
ctx, cancel := context.WithCancel(context.Background())
sender := newPipelinedSender(ctx, stream, 1, false, "local")
msg := &filer_pb.SubscribeMetadataResponse{EventNotification: &filer_pb.EventNotification{}}
if err := sender.Send(msg); err != nil { // consumed by the blocked sendLoop
t.Fatal(err)
}
if err := sender.Send(msg); err != nil { // fills the one-deep queue
t.Fatal(err)
}
cancel()
done := make(chan error, 1)
go func() { done <- sender.Send(msg) }()
select {
case err := <-done:
if err != context.Canceled {
t.Fatalf("Send returned %v, want context.Canceled", err)
}
case <-time.After(2 * time.Second):
t.Fatal("Send still blocked after cancel")
}
if err := sender.Close(); err != nil {
t.Fatalf("Close returned %v", err)
}
}
// Once the subscription context ends, sendLoop must not start another
// stream.Send: the handler returns and gRPC ends the stream behind it, so a
// send started then would race the stream's teardown. The one Send already
// in flight is released by that teardown.
func TestPipelinedSenderStopsSendingOnCancel(t *testing.T) {
stream := &blockedStream{release: make(chan struct{}), entered: make(chan struct{}, 1)}
ctx, cancel := context.WithCancel(context.Background())
sender := newPipelinedSender(ctx, stream, 4, false, "local")
msg := &filer_pb.SubscribeMetadataResponse{EventNotification: &filer_pb.EventNotification{}}
if err := sender.Send(msg); err != nil {
t.Fatal(err)
}
<-stream.entered // sendLoop is inside stream.Send
for i := 0; i < 2; i++ { // queued behind the blocked send
if err := sender.Send(msg); err != nil {
t.Fatal(err)
}
}
cancel()
close(stream.release)
if err := sender.Close(); err != nil {
t.Fatalf("Close returned %v", err)
}
select {
case <-sender.done:
case <-time.After(2 * time.Second):
t.Fatal("sendLoop did not exit after cancel")
}
if n := stream.sends.Load(); n != 1 {
t.Fatalf("stream.Send ran %d times, want 1; a new send started after cancel", n)
}
}
// deliveryLagSnapshot returns the delivery-lag histogram's sample count and sum
// for a scope, so a test can assert what one send added.
func deliveryLagSnapshot(t *testing.T, scope string) (uint64, float64) {
t.Helper()
var m dto.Metric
if err := stats.FilerSubscribeDeliveryLagSeconds.WithLabelValues(scope).(prometheus.Metric).Write(&m); err != nil {
t.Fatalf("read delivery-lag metric: %v", err)
}
return m.GetHistogram().GetSampleCount(), m.GetHistogram().GetSampleSum()
}
// gatedStream blocks Send until release is closed, signalling on entered the
// first time Send is called, so a test can inspect state while a send is in
// flight (i.e. before it has succeeded).
type gatedStream struct {
recordingStream
entered chan struct{}
release chan struct{}
}
func (s *gatedStream) Send(m *filer_pb.SubscribeMetadataResponse) error {
select {
case s.entered <- struct{}{}:
default:
}
<-s.release
return s.recordingStream.Send(m)
}
// TestPipelinedSenderObservesDeliveryLagAfterSend proves delivery lag is sampled
// after the real stream send (through the queue + Close), measured from the
// event's mutation stamp (not the moved stream position), and that control
// messages and empty progress markers are not sampled.
func TestPipelinedSenderObservesDeliveryLagAfterSend(t *testing.T) {
const scope = "delivery-lag-test"
beforeCount, beforeSum := deliveryLagSnapshot(t, scope)
stream := &gatedStream{entered: make(chan struct{}, 1), release: make(chan struct{})}
sender := newPipelinedSender(context.Background(), stream, 64, false, scope)
mutationNs := time.Now().Add(-5 * time.Second).UnixNano()
// A real event whose stream position (TsNs) has moved well past its original
// mutation stamp; lag must be measured from the mutation stamp.
event := &filer_pb.SubscribeMetadataResponse{
TsNs: time.Now().UnixNano(),
MutationTsNs: mutationNs,
EventNotification: &filer_pb.EventNotification{NewEntry: &filer_pb.Entry{Name: "e"}},
}
// A control message (heartbeat): no EventNotification -> must not be sampled.
control := &filer_pb.SubscribeMetadataResponse{TsNs: time.Now().UnixNano()}
// A progress marker (sent after MaxUnsyncedEvents filtered entries): a
// non-nil but empty EventNotification carrying the filtered entry's TsNs ->
// must not be sampled either.
marker := &filer_pb.SubscribeMetadataResponse{
TsNs: time.Now().UnixNano(),
EventNotification: &filer_pb.EventNotification{},
}
if err := sender.Send(event); err != nil {
t.Fatalf("send event: %v", err)
}
// Wait until the send loop is blocked inside stream.Send for the event, then
// confirm nothing was sampled yet: the observation must come after the send
// succeeds, not when the event entered the queue.
<-stream.entered
if midCount, _ := deliveryLagSnapshot(t, scope); midCount != beforeCount {
t.Fatalf("delivery-lag sampled %d while the send was still blocked; it must observe only after stream.Send succeeds", midCount-beforeCount)
}
close(stream.release)
if err := sender.Send(control); err != nil {
t.Fatalf("send control: %v", err)
}
if err := sender.Send(marker); err != nil {
t.Fatalf("send marker: %v", err)
}
if err := sender.Close(); err != nil {
t.Fatalf("close: %v", err)
}
afterCount, afterSum := deliveryLagSnapshot(t, scope)
if afterCount != beforeCount+1 {
t.Fatalf("delivery-lag samples added = %d, want 1 (only the real event; control and progress markers skipped)", afterCount-beforeCount)
}
if lag := afterSum - beforeSum; lag < 4 || lag > 60 {
t.Fatalf("delivery lag = %.2fs, want ~5s from the mutation stamp (not ~0 from the moved TsNs)", lag)
}
}