perf(chatd): fix six scale bottlenecks identified by benchmarking (#22957)

## Summary

Scale-tested the `chatd` package with mock-based benchmarks to identify
performance bottlenecks. This PR fixes 6 of the 8 identified issues,
ranked by severity.

## Changes

### 1. Parallel tool execution (HIGH) — `chatloop.go`
`executeTools` ran tool calls sequentially. Now dispatches all calls
concurrently via goroutines with `sync.WaitGroup`. Results are
pre-allocated by index (no mutex needed). `onResult` callbacks fire as
each tool completes.

### 2. Pubsub-backed subagent await (HIGH) — `subagent.go`
`awaitSubagentCompletion` polled the DB every 200ms. Now subscribes to
the child chat's `ChatStreamNotifyChannel` via pubsub for near-instant
notifications. Fallback poll reduced to 5s. Falls back to 200ms only
when `pubsub == nil` (single-instance / in-memory).

### 3. Per-chat stream locking (MEDIUM) — `chatd.go`
Replaced single global `streamMu` + `map[uuid.UUID]*chatStreamState`
with `sync.Map` where each `chatStreamState` has its own `sync.Mutex`.
Zero cross-chat contention.

### 4. Batch chat acquisition (MEDIUM) — `chatd.go`
`processOnce` acquired 1 chat per tick. Now loops up to
`maxChatsPerAcquire = 10` per tick, avoiding idle time when many chats
are pending.

### 5. Reduced heartbeat frequency (LOW-MEDIUM) — `chatd.go`
`chatHeartbeatInterval` changed from 30s to 60s. Safe given the 5-minute
`DefaultInFlightChatStaleAfter`.

### 6. O(depth) descendant check (LOW) — `subagent.go`
Replaced top-down BFS (`O(total_descendants)` queries) with bottom-up
parent-chain walk (`O(depth)` queries). Includes cycle protection.

## Not addressed (intentionally)
- Message serialization overhead
- Buffer eviction (`buffer[1:]` pattern)
This commit is contained in:
Kyle Carberry
2026-03-11 14:00:08 -04:00
committed by GitHub
parent e5c19d0af4
commit 1f37df4db3
10 changed files with 304 additions and 223 deletions
+4 -4
View File
@@ -1512,13 +1512,13 @@ func (q *querier) authorizeProvisionerJob(ctx context.Context, job database.Prov
return nil
}
func (q *querier) AcquireChat(ctx context.Context, arg database.AcquireChatParams) (database.Chat, error) {
// AcquireChat is a system-level operation used by the chat processor.
func (q *querier) AcquireChats(ctx context.Context, arg database.AcquireChatsParams) ([]database.Chat, error) {
// AcquireChats is a system-level operation used by the chat processor.
// Authorization is done at the system level, not per-user.
if err := q.authorizeContext(ctx, policy.ActionUpdate, rbac.ResourceChat); err != nil {
return database.Chat{}, err
return nil, err
}
return q.db.AcquireChat(ctx, arg)
return q.db.AcquireChats(ctx, arg)
}
func (q *querier) AcquireLock(ctx context.Context, id int64) error {
+5 -4
View File
@@ -373,14 +373,15 @@ func (s *MethodTestSuite) TestConnectionLogs() {
}
func (s *MethodTestSuite) TestChats() {
s.Run("AcquireChat", s.Mocked(func(dbm *dbmock.MockStore, faker *gofakeit.Faker, check *expects) {
arg := database.AcquireChatParams{
s.Run("AcquireChats", s.Mocked(func(dbm *dbmock.MockStore, faker *gofakeit.Faker, check *expects) {
arg := database.AcquireChatsParams{
StartedAt: dbtime.Now(),
WorkerID: uuid.New(),
NumChats: 1,
}
chat := testutil.Fake(s.T(), faker, database.Chat{})
dbm.EXPECT().AcquireChat(gomock.Any(), arg).Return(chat, nil).AnyTimes()
check.Args(arg).Asserts(rbac.ResourceChat, policy.ActionUpdate).Returns(chat)
dbm.EXPECT().AcquireChats(gomock.Any(), arg).Return([]database.Chat{chat}, nil).AnyTimes()
check.Args(arg).Asserts(rbac.ResourceChat, policy.ActionUpdate).Returns([]database.Chat{chat})
}))
s.Run("DeleteAllChatQueuedMessages", s.Mocked(func(dbm *dbmock.MockStore, faker *gofakeit.Faker, check *expects) {
chat := testutil.Fake(s.T(), faker, database.Chat{})
+4 -4
View File
@@ -104,11 +104,11 @@ func (m queryMetricsStore) DeleteOrganization(ctx context.Context, id uuid.UUID)
return r0
}
func (m queryMetricsStore) AcquireChat(ctx context.Context, arg database.AcquireChatParams) (database.Chat, error) {
func (m queryMetricsStore) AcquireChats(ctx context.Context, arg database.AcquireChatsParams) ([]database.Chat, error) {
start := time.Now()
r0, r1 := m.s.AcquireChat(ctx, arg)
m.queryLatencies.WithLabelValues("AcquireChat").Observe(time.Since(start).Seconds())
m.queryCounts.WithLabelValues(httpmw.ExtractHTTPRoute(ctx), httpmw.ExtractHTTPMethod(ctx), "AcquireChat").Inc()
r0, r1 := m.s.AcquireChats(ctx, arg)
m.queryLatencies.WithLabelValues("AcquireChats").Observe(time.Since(start).Seconds())
m.queryCounts.WithLabelValues(httpmw.ExtractHTTPRoute(ctx), httpmw.ExtractHTTPMethod(ctx), "AcquireChats").Inc()
return r0, r1
}
+7 -7
View File
@@ -44,19 +44,19 @@ func (m *MockStore) EXPECT() *MockStoreMockRecorder {
return m.recorder
}
// AcquireChat mocks base method.
func (m *MockStore) AcquireChat(ctx context.Context, arg database.AcquireChatParams) (database.Chat, error) {
// AcquireChats mocks base method.
func (m *MockStore) AcquireChats(ctx context.Context, arg database.AcquireChatsParams) ([]database.Chat, error) {
m.ctrl.T.Helper()
ret := m.ctrl.Call(m, "AcquireChat", ctx, arg)
ret0, _ := ret[0].(database.Chat)
ret := m.ctrl.Call(m, "AcquireChats", ctx, arg)
ret0, _ := ret[0].([]database.Chat)
ret1, _ := ret[1].(error)
return ret0, ret1
}
// AcquireChat indicates an expected call of AcquireChat.
func (mr *MockStoreMockRecorder) AcquireChat(ctx, arg any) *gomock.Call {
// AcquireChats indicates an expected call of AcquireChats.
func (mr *MockStoreMockRecorder) AcquireChats(ctx, arg any) *gomock.Call {
mr.mock.ctrl.T.Helper()
return mr.mock.ctrl.RecordCallWithMethodType(mr.mock, "AcquireChat", reflect.TypeOf((*MockStore)(nil).AcquireChat), ctx, arg)
return mr.mock.ctrl.RecordCallWithMethodType(mr.mock, "AcquireChats", reflect.TypeOf((*MockStore)(nil).AcquireChats), ctx, arg)
}
// AcquireLock mocks base method.
+3 -3
View File
@@ -12,9 +12,9 @@ import (
)
type sqlcQuerier interface {
// Acquires a pending chat for processing. Uses SKIP LOCKED to prevent
// multiple replicas from acquiring the same chat.
AcquireChat(ctx context.Context, arg AcquireChatParams) (Chat, error)
// Acquires up to @num_chats pending chats for processing. Uses SKIP LOCKED
// to prevent multiple replicas from acquiring the same chat.
AcquireChats(ctx context.Context, arg AcquireChatsParams) ([]Chat, error)
// Blocks until the lock is acquired.
//
// This must be called from within a transaction. The lock will be automatically
+44 -27
View File
@@ -2968,7 +2968,7 @@ func (q *sqlQuerier) UpdateChatProvider(ctx context.Context, arg UpdateChatProvi
return i, err
}
const acquireChat = `-- name: AcquireChat :one
const acquireChats = `-- name: AcquireChats :many
UPDATE
chats
SET
@@ -2978,7 +2978,7 @@ SET
updated_at = $1::timestamptz,
worker_id = $2::uuid
WHERE
id = (
id = ANY(
SELECT
id
FROM
@@ -2990,40 +2990,57 @@ WHERE
FOR UPDATE
SKIP LOCKED
LIMIT
1
$3::int
)
RETURNING
id, owner_id, workspace_id, title, status, worker_id, started_at, heartbeat_at, created_at, updated_at, parent_chat_id, root_chat_id, last_model_config_id, archived, last_error
`
type AcquireChatParams struct {
type AcquireChatsParams struct {
StartedAt time.Time `db:"started_at" json:"started_at"`
WorkerID uuid.UUID `db:"worker_id" json:"worker_id"`
NumChats int32 `db:"num_chats" json:"num_chats"`
}
// Acquires a pending chat for processing. Uses SKIP LOCKED to prevent
// multiple replicas from acquiring the same chat.
func (q *sqlQuerier) AcquireChat(ctx context.Context, arg AcquireChatParams) (Chat, error) {
row := q.db.QueryRowContext(ctx, acquireChat, arg.StartedAt, arg.WorkerID)
var i Chat
err := row.Scan(
&i.ID,
&i.OwnerID,
&i.WorkspaceID,
&i.Title,
&i.Status,
&i.WorkerID,
&i.StartedAt,
&i.HeartbeatAt,
&i.CreatedAt,
&i.UpdatedAt,
&i.ParentChatID,
&i.RootChatID,
&i.LastModelConfigID,
&i.Archived,
&i.LastError,
)
return i, err
// Acquires up to @num_chats pending chats for processing. Uses SKIP LOCKED
// to prevent multiple replicas from acquiring the same chat.
func (q *sqlQuerier) AcquireChats(ctx context.Context, arg AcquireChatsParams) ([]Chat, error) {
rows, err := q.db.QueryContext(ctx, acquireChats, arg.StartedAt, arg.WorkerID, arg.NumChats)
if err != nil {
return nil, err
}
defer rows.Close()
var items []Chat
for rows.Next() {
var i Chat
if err := rows.Scan(
&i.ID,
&i.OwnerID,
&i.WorkspaceID,
&i.Title,
&i.Status,
&i.WorkerID,
&i.StartedAt,
&i.HeartbeatAt,
&i.CreatedAt,
&i.UpdatedAt,
&i.ParentChatID,
&i.RootChatID,
&i.LastModelConfigID,
&i.Archived,
&i.LastError,
); err != nil {
return nil, err
}
items = append(items, i)
}
if err := rows.Close(); err != nil {
return nil, err
}
if err := rows.Err(); err != nil {
return nil, err
}
return items, nil
}
const acquireStaleChatDiffStatuses = `-- name: AcquireStaleChatDiffStatuses :many
+5 -5
View File
@@ -257,9 +257,9 @@ WHERE
RETURNING
*;
-- name: AcquireChat :one
-- Acquires a pending chat for processing. Uses SKIP LOCKED to prevent
-- multiple replicas from acquiring the same chat.
-- name: AcquireChats :many
-- Acquires up to @num_chats pending chats for processing. Uses SKIP LOCKED
-- to prevent multiple replicas from acquiring the same chat.
UPDATE
chats
SET
@@ -269,7 +269,7 @@ SET
updated_at = @started_at::timestamptz,
worker_id = @worker_id::uuid
WHERE
id = (
id = ANY(
SELECT
id
FROM
@@ -281,7 +281,7 @@ WHERE
FOR UPDATE
SKIP LOCKED
LIMIT
1
@num_chats::int
)
RETURNING
*;