Skip to content

refactor: separate pub/sub interface from durable task queue #4480

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Merged
igor-kupczynski merged 7 commits into from
Jul 28, 2026
Merged

refactor: separate pub/sub interface from durable task queue #4480

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
7 commits
Select commit Hold shift + click to select a range
d239557
refactor: pub/sub separate from durable task queue
abelanger5 Jul 22, 2026
69c3092
address review: topic kind enum, MsgHandler rename, PubTenantMessage …
abelanger5 Jul 22, 2026
3c8da09
Merge remote-tracking branch 'origin/main' into belanger/msgqueue-pub…
abelanger5 Jul 22, 2026
a825a7b
address review: drop pubsub from health probes, decompress cross-back…
abelanger5 Jul 27, 2026
d199b35
address review: move Compressor to msgqueue, Message.Clone, real post…
abelanger5 Jul 27, 2026
ebafd2f
fix TestMsgIdBufferMemoryLeak failing after new tests joined the package
abelanger5 Jul 27, 2026
8fdbad1
address review: reset Compressed after rabbitmq pubsub decompress, do…
igor-kupczynski Jul 28, 2026
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
14 changes: 14 additions & 0 deletions cmd/hatchet-engine/engine/run.go
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -197,6 +197,7 @@ func runV0Config(ctx context.Context, sc *server.ServerConfig, cleanup *cleanup.
sv1, err := schedulerv1.New(
schedulerv1.WithAlerter(sc.Alerter),
schedulerv1.WithMessageQueue(sc.MessageQueueV1),
schedulerv1.WithPubSub(sc.PubSubV1),
schedulerv1.WithRepository(sc.V1),
schedulerv1.WithLogger(sc.Logger),
schedulerv1.WithPartition(p),
Expand Down Expand Up @@ -249,6 +250,7 @@ func runV0Config(ctx context.Context, sc *server.ServerConfig, cleanup *cleanup.
tasks, err := task.New(
task.WithAlerter(sc.Alerter),
task.WithMessageQueue(sc.MessageQueueV1),
task.WithPubSub(sc.PubSubV1),
task.WithV1Repository(sc.V1),
task.WithLogger(sc.Logger),
task.WithPartition(p),
Expand Down Expand Up @@ -281,6 +283,7 @@ func runV0Config(ctx context.Context, sc *server.ServerConfig, cleanup *cleanup.
olap, err := olap.New(
olap.WithAlerter(sc.Alerter),
olap.WithMessageQueue(sc.MessageQueueV1),
olap.WithPubSub(sc.PubSubV1),
olap.WithRepository(sc.V1),
olap.WithLogger(sc.Logger),
olap.WithPartition(p),
Expand Down Expand Up @@ -343,6 +346,7 @@ func runV0Config(ctx context.Context, sc *server.ServerConfig, cleanup *cleanup.
d, err := dispatcher.New(
dispatcher.WithAlerter(sc.Alerter),
dispatcher.WithMessageQueueV1(sc.MessageQueueV1),
dispatcher.WithPubSub(sc.PubSubV1),
dispatcher.WithRepositoryV1(sc.V1),
dispatcher.WithLogger(sc.Logger),
dispatcher.WithCache(cacheInstance),
Expand All @@ -367,6 +371,7 @@ func runV0Config(ctx context.Context, sc *server.ServerConfig, cleanup *cleanup.
// create the event ingestor
ei, err := ingestor.NewIngestor(
ingestor.WithMessageQueueV1(sc.MessageQueueV1),
ingestor.WithPubSub(sc.PubSubV1),
ingestor.WithRepositoryV1(sc.V1),
ingestor.WithLogIngestionEnabled(sc.Runtime.LogIngestionEnabled),
ingestor.WithLocalScheduler(localScheduler),
Expand All @@ -385,6 +390,7 @@ func runV0Config(ctx context.Context, sc *server.ServerConfig, cleanup *cleanup.
adminSvc, err := admin.NewAdminService(
admin.WithRepositoryV1(sc.V1),
admin.WithMessageQueueV1(sc.MessageQueueV1),
admin.WithPubSub(sc.PubSubV1),
admin.WithLocalScheduler(localScheduler),
admin.WithLocalDispatcher(d),
admin.WithOptimisticSchedulingEnabled(sc.Runtime.OptimisticSchedulingEnabled),
Expand All @@ -400,6 +406,7 @@ func runV0Config(ctx context.Context, sc *server.ServerConfig, cleanup *cleanup.
adminv1Svc, err := adminv1.NewAdminService(
adminv1.WithRepository(sc.V1),
adminv1.WithMessageQueue(sc.MessageQueueV1),
adminv1.WithPubSub(sc.PubSubV1),
adminv1.WithAnalytics(sc.Analytics),
adminv1.WithLocalScheduler(localScheduler),
adminv1.WithLocalDispatcher(d),
Expand Down Expand Up @@ -631,6 +638,7 @@ func runV1Config(ctx context.Context, sc *server.ServerConfig, cleanup *cleanup.
tasks, err := task.New(
task.WithAlerter(sc.Alerter),
task.WithMessageQueue(sc.MessageQueueV1),
task.WithPubSub(sc.PubSubV1),
task.WithV1Repository(sc.V1),
task.WithLogger(sc.Logger),
task.WithPartition(p),
Expand Down Expand Up @@ -666,6 +674,7 @@ func runV1Config(ctx context.Context, sc *server.ServerConfig, cleanup *cleanup.
olap, err := olap.New(
olap.WithAlerter(sc.Alerter),
olap.WithMessageQueue(sc.MessageQueueV1),
olap.WithPubSub(sc.PubSubV1),
olap.WithRepository(sc.V1),
olap.WithLogger(sc.Logger),
olap.WithPartition(p),
Expand Down Expand Up @@ -748,6 +757,7 @@ func runV1Config(ctx context.Context, sc *server.ServerConfig, cleanup *cleanup.
sv1, err := schedulerv1.New(
schedulerv1.WithAlerter(sc.Alerter),
schedulerv1.WithMessageQueue(sc.MessageQueueV1),
schedulerv1.WithPubSub(sc.PubSubV1),
schedulerv1.WithRepository(sc.V1),
schedulerv1.WithLogger(sc.Logger),
schedulerv1.WithPartition(p),
Expand Down Expand Up @@ -781,6 +791,7 @@ func runV1Config(ctx context.Context, sc *server.ServerConfig, cleanup *cleanup.
d, err := dispatcher.New(
dispatcher.WithAlerter(sc.Alerter),
dispatcher.WithMessageQueueV1(sc.MessageQueueV1),
dispatcher.WithPubSub(sc.PubSubV1),
dispatcher.WithRepositoryV1(sc.V1),
dispatcher.WithLogger(sc.Logger),
dispatcher.WithCache(cacheInstance),
Expand All @@ -806,6 +817,7 @@ func runV1Config(ctx context.Context, sc *server.ServerConfig, cleanup *cleanup.
// create the event ingestor
ei, err := ingestor.NewIngestor(
ingestor.WithMessageQueueV1(sc.MessageQueueV1),
ingestor.WithPubSub(sc.PubSubV1),
ingestor.WithRepositoryV1(sc.V1),
ingestor.WithLogIngestionEnabled(sc.Runtime.LogIngestionEnabled),
ingestor.WithLocalScheduler(localScheduler),
Expand All @@ -824,6 +836,7 @@ func runV1Config(ctx context.Context, sc *server.ServerConfig, cleanup *cleanup.
adminSvc, err := admin.NewAdminService(
admin.WithRepositoryV1(sc.V1),
admin.WithMessageQueueV1(sc.MessageQueueV1),
admin.WithPubSub(sc.PubSubV1),
admin.WithLocalScheduler(localScheduler),
admin.WithLocalDispatcher(d),
admin.WithOptimisticSchedulingEnabled(sc.Runtime.OptimisticSchedulingEnabled),
Expand All @@ -840,6 +853,7 @@ func runV1Config(ctx context.Context, sc *server.ServerConfig, cleanup *cleanup.
adminv1Svc, err := adminv1.NewAdminService(
adminv1.WithRepository(sc.V1),
adminv1.WithMessageQueue(sc.MessageQueueV1),
adminv1.WithPubSub(sc.PubSubV1),
adminv1.WithAnalytics(sc.Analytics),
adminv1.WithLocalScheduler(localScheduler),
adminv1.WithLocalDispatcher(d),
Expand Down
30 changes: 18 additions & 12 deletions frontend/docs/pages/self-hosting/configuration-options.mdx
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -291,18 +291,24 @@ Variables marked with ⚠️ are conditionally required when specific features a

## Task Queue Configuration

| Variable | Description | Default Value |
| --------------------------------------------------- | -------------------------------------------------------------- | ------------- |
| `SERVER_MSGQUEUE_KIND` | Message queue kind (`rabbitmq` or `postgres`) | `rabbitmq` |
| `SERVER_MSGQUEUE_RABBITMQ_URL` | RabbitMQ URL | |
| `SERVER_MSGQUEUE_RABBITMQ_QOS` | RabbitMQ QoS (prefetch count) | `100` |
| `SERVER_MSGQUEUE_RABBITMQ_MAX_PUB_CHANS` | Max RabbitMQ publish channels | `20` |
| `SERVER_MSGQUEUE_RABBITMQ_MAX_SUB_CHANS` | Max RabbitMQ subscribe channels | `100` |
| `SERVER_MSGQUEUE_RABBITMQ_COMPRESSION_ENABLED` | Enable gzip compression of messages | `false` |
| `SERVER_MSGQUEUE_RABBITMQ_COMPRESSION_THRESHOLD` | Byte size above which messages are compressed | `5120` |
| `SERVER_MSGQUEUE_RABBITMQ_ENABLE_MESSAGE_REJECTION` | Reject (rather than requeue) messages past the max death count | `false` |
| `SERVER_MSGQUEUE_RABBITMQ_MAX_DEATH_COUNT` | Max redeliveries before a message is dead-lettered | `1000` |
| `SERVER_SINGLE_QUEUE_LIMIT` | Single queue limit | `100` |
| Variable | Description | Default Value |
| --------------------------------------------------- | ---------------------------------------------------------------------------------------------------------- | ------------- |
| `SERVER_MSGQUEUE_KIND` | Message queue kind (`rabbitmq` or `postgres`) | `rabbitmq` |
| `SERVER_MSGQUEUE_RABBITMQ_URL` | RabbitMQ URL | |
| `SERVER_MSGQUEUE_RABBITMQ_QOS` | RabbitMQ QoS (prefetch count) | `100` |
| `SERVER_MSGQUEUE_RABBITMQ_MAX_PUB_CHANS` | Max RabbitMQ publish channels | `20` |
| `SERVER_MSGQUEUE_RABBITMQ_MAX_SUB_CHANS` | Max RabbitMQ subscribe channels | `100` |
| `SERVER_MSGQUEUE_RABBITMQ_COMPRESSION_ENABLED` | Enable gzip compression of messages | `false` |
| `SERVER_MSGQUEUE_RABBITMQ_COMPRESSION_THRESHOLD` | Byte size above which messages are compressed | `5120` |
| `SERVER_MSGQUEUE_RABBITMQ_ENABLE_MESSAGE_REJECTION` | Reject (rather than requeue) messages past the max death count | `false` |
| `SERVER_MSGQUEUE_RABBITMQ_MAX_DEATH_COUNT` | Max redeliveries before a message is dead-lettered | `1000` |
| `SERVER_MSGQUEUE_PUBSUB_KIND` | Pub/sub kind for best-effort notifications (inherits `SERVER_MSGQUEUE_KIND` when unset) | |
| `SERVER_MSGQUEUE_PUBSUB_RABBITMQ_URL` | Pub/sub RabbitMQ URL (inherits `SERVER_MSGQUEUE_RABBITMQ_URL` when unset; always uses its own connections) | |
| `SERVER_MSGQUEUE_PUBSUB_RABBITMQ_MAX_PUB_CHANS` | Pub/sub RabbitMQ max publish channels | `10` |
| `SERVER_MSGQUEUE_PUBSUB_RABBITMQ_MAX_SUB_CHANS` | Pub/sub RabbitMQ max subscribe channels | `20` |
| `SERVER_MSGQUEUE_PUBSUB_POSTGRES_MAX_CONNS` | Pub/sub Postgres pool max connections (pool is built from the direct `DATABASE_URL`) | `5` |
| `SERVER_MSGQUEUE_PUBSUB_POSTGRES_MIN_CONNS` | Pub/sub Postgres pool min connections | `1` |
| `SERVER_SINGLE_QUEUE_LIMIT` | Single queue limit | `100` |

### Message Queue Buffers

Expand Down
36 changes: 26 additions & 10 deletions internal/msgqueue/rabbitmq/gzip.go → internal/msgqueue/gzip.go
View file
Open in desktop
Original file line number Diff line number Diff line change
@@ -1,14 +1,25 @@
package rabbitmq
package msgqueue

import (
"bytes"
"compress/gzip"
"fmt"
"io"

"sync"
)

// DefaultCompressionThreshold is the payload size above which messages are
// gzip-compressed when compression is enabled.
const DefaultCompressionThreshold = 5 * 1024 // 5KB

// Compressor holds gzip compression settings for a message queue
// implementation. Compression settings must agree between the durable queue
// and the pub/sub, since both publish to the same tenant topics.
type Compressor struct {
Enabled bool
Threshold int
}

type CompressionResult struct {
Payloads [][]byte
WasCompressed bool
Expand All @@ -23,7 +34,7 @@ type CompressionResult struct {
// No explicit size cap is needed: sync.Pool is self-limiting because the Go
// runtime evicts pooled objects during GC, so the pool cannot grow unbounded.
// In practice the pool size is also bounded by the number of goroutines
// concurrently compressing, which is small for a RabbitMQ publish path.
// concurrently compressing, which is small for a publish path.
var gzipWriterPool = sync.Pool{
New: func() any {
return gzip.NewWriter(nil)
Expand All @@ -38,15 +49,16 @@ func getPayloadSize(payloads [][]byte) int {
return totalSize
}

// compressPayloads compresses message payloads using gzip if they exceed the minimum size threshold.
// Returns compression results including the compressed payloads and compression statistics.
func (t *MessageQueueImpl) compressPayloads(payloads [][]byte) (*CompressionResult, error) {
// CompressPayloads compresses message payloads using gzip if they exceed the
// minimum size threshold. Returns compression results including the compressed
// payloads and compression statistics.
func (t Compressor) CompressPayloads(payloads [][]byte) (*CompressionResult, error) {
result := &CompressionResult{
Payloads: payloads,
WasCompressed: false,
}

if !t.compressionEnabled || len(payloads) == 0 {
if !t.Enabled || len(payloads) == 0 {
return result, nil
}

Expand All @@ -55,7 +67,7 @@ func (t *MessageQueueImpl) compressPayloads(payloads [][]byte) (*CompressionResu
result.OriginalSize = totalSize

// Only compress if total size exceeds threshold
if totalSize < t.compressionThreshold {
if totalSize < t.Threshold {
result.CompressedSize = totalSize
result.CompressionRatio = 1.0
return result, nil
Expand Down Expand Up @@ -97,8 +109,12 @@ func (t *MessageQueueImpl) compressPayloads(payloads [][]byte) (*CompressionResu
return result, nil
}

// decompressPayloads decompresses message payloads using gzip.
func (t *MessageQueueImpl) decompressPayloads(payloads [][]byte) ([][]byte, error) {
// DecompressPayloads decompresses gzip-compressed message payloads. It lives
// in the msgqueue package (rather than a single implementation) because a
// compressed message can cross backends: a producer whose durable queue is
// RabbitMQ with compression enabled hands the same *Message to whichever
// pub/sub is configured, so every subscriber must be able to decompress.
func DecompressPayloads(payloads [][]byte) ([][]byte, error) {
if len(payloads) == 0 {
return payloads, nil
}
Expand Down
48 changes: 24 additions & 24 deletions internal/msgqueue/rabbitmq/gzip_test.go → internal/msgqueue/gzip_test.go
View file
Open in desktop
Original file line number Diff line number Diff line change
@@ -1,4 +1,4 @@
package rabbitmq
package msgqueue

import (
"bytes"
Expand All @@ -25,23 +25,23 @@ func generatePayloads(count, size int) [][]byte {
return payloads
}

func newMQ() *MessageQueueImpl {
return &MessageQueueImpl{
compressionEnabled: true,
compressionThreshold: 0,
func newCompressor() Compressor {
return Compressor{
Enabled: true,
Threshold: 0,
}
}

func TestCompressDecompressRoundtrip(t *testing.T) {
mq := newMQ()
mq := newCompressor()
payloads := generatePayloads(5, 10*1024)
result, err := mq.compressPayloads(payloads)
result, err := mq.CompressPayloads(payloads)

assert.NoError(t, err)
assert.True(t, result.WasCompressed, "expected WasCompressed to be true")
assert.Equal(t, len(payloads), len(result.Payloads), "expected %d payloads, got %d", len(payloads), len(result.Payloads))

decompressed, err := mq.decompressPayloads(result.Payloads)
decompressed, err := DecompressPayloads(result.Payloads)

assert.NoError(t, err)

Expand All @@ -52,69 +52,69 @@ func TestCompressDecompressRoundtrip(t *testing.T) {
}

func TestCompressPayloadsDisabled(t *testing.T) {
mq := &MessageQueueImpl{
compressionEnabled: false,
compressionThreshold: 0,
mq := Compressor{
Enabled: false,
Threshold: 0,
}

payloads := generatePayloads(3, 1024)
result, err := mq.compressPayloads(payloads)
result, err := mq.CompressPayloads(payloads)

assert.NoError(t, err)
assert.False(t, result.WasCompressed, "expected WasCompressed to be false when compression is disabled")
}

func TestCompressPayloadsBelowThreshold(t *testing.T) {
mq := &MessageQueueImpl{
compressionEnabled: true,
compressionThreshold: 100 * 1024,
mq := Compressor{
Enabled: true,
Threshold: 100 * 1024,
}

payloads := generatePayloads(1, 1024)
result, err := mq.compressPayloads(payloads)
result, err := mq.CompressPayloads(payloads)

assert.NoError(t, err)
assert.False(t, result.WasCompressed, "expected WasCompressed to be false when below threshold")
}

func BenchmarkCompressPayloads_1x10KiB(b *testing.B) {
mq := newMQ()
mq := newCompressor()
payloads := generatePayloads(1, 10*1024)
b.ResetTimer()
b.ReportAllocs()
for i := 0; i < b.N; i++ {
_, _ = mq.compressPayloads(payloads)
_, _ = mq.CompressPayloads(payloads)
}
}

func BenchmarkCompressPayloads_10x10KiB(b *testing.B) {
mq := newMQ()
mq := newCompressor()
payloads := generatePayloads(10, 10*1024)
b.ResetTimer()
b.ReportAllocs()
for i := 0; i < b.N; i++ {
_, _ = mq.compressPayloads(payloads)
_, _ = mq.CompressPayloads(payloads)
}
}

func BenchmarkCompressPayloads_10x100KiB(b *testing.B) {
mq := newMQ()
mq := newCompressor()
payloads := generatePayloads(10, 100*1024)
b.ResetTimer()
b.ReportAllocs()
for i := 0; i < b.N; i++ {
_, _ = mq.compressPayloads(payloads)
_, _ = mq.CompressPayloads(payloads)
}
}

func BenchmarkCompressPayloads_Concurrent(b *testing.B) {
mq := newMQ()
mq := newCompressor()
payloads := generatePayloads(5, 10*1024)
b.ResetTimer()
b.ReportAllocs()
b.RunParallel(func(pb *testing.PB) {
for pb.Next() {
_, _ = mq.compressPayloads(payloads)
_, _ = mq.CompressPayloads(payloads)
}
})
}
7 changes: 2 additions & 5 deletions internal/msgqueue/mq_pub_buffer_test.go
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -5,8 +5,6 @@ import (
"sync/atomic"
"testing"
"time"

"github.com/google/uuid"
)

// mockMessageQueue satisfies MessageQueue for tests. Only SendMessage is wired up.
Expand All @@ -24,11 +22,10 @@ func (m *mockMessageQueue) SendMessage(ctx context.Context, q Queue, msg *Messag
}
return nil
}
func (m *mockMessageQueue) Subscribe(_ Queue, _ AckHook, _ AckHook) (func() error, error) {
func (m *mockMessageQueue) Subscribe(_ Queue, _ MsgHandler, _ MsgHandler) (func() error, error) {
return func() error { return nil }, nil
}
func (m *mockMessageQueue) RegisterTenant(_ context.Context, _ uuid.UUID) error { return nil }
func (m *mockMessageQueue) IsReady() bool { return true }
func (m *mockMessageQueue) IsReady() bool { return true }

// TestPubBufferFlushesWhenFull verifies that when the channel is at capacity the
// capacityRelease mechanism breaks the post-flush interval wait and triggers an
Expand Down
Loading
Loading