Add resource limits to daemon (bd-152)

Implemented connection limiting, request timeouts, and memory pressure detection:

- Connection limiting with semaphore pattern (default 100 max connections)
- Request timeout enforcement on read/write (default 30s)
- Memory pressure detection with aggressive cache eviction (default 500MB threshold)
- Configurable via environment variables:
  - BEADS_DAEMON_MAX_CONNS
  - BEADS_DAEMON_REQUEST_TIMEOUT
  - BEADS_DAEMON_MEMORY_THRESHOLD_MB
- Health endpoint now exposes active/max connections and memory usage
- Comprehensive test coverage for all limits

This prevents resource exhaustion under heavy load or attack scenarios.

Amp-Thread-ID: https://ampcode.com/threads/T-44d1817a-3709-4f1d-a27a-78bb2fa4d3dc
Co-authored-by: Amp <amp@ampcode.com>
This commit is contained in:
Steve Yegge
2025-10-19 13:22:23 -07:00
parent 73cc958ecd
commit c71da4c267
4 changed files with 464 additions and 10 deletions

330
internal/rpc/limits_test.go Normal file
View File

@@ -0,0 +1,330 @@
package rpc
import (
"bufio"
"context"
"encoding/json"
"fmt"
"net"
"os"
"path/filepath"
"sync"
"sync/atomic"
"testing"
"time"
"github.com/steveyegge/beads/internal/storage/sqlite"
)
func TestConnectionLimits(t *testing.T) {
tmpDir := t.TempDir()
dbPath := filepath.Join(tmpDir, ".beads", "test.db")
if err := os.MkdirAll(filepath.Dir(dbPath), 0755); err != nil {
t.Fatal(err)
}
store, err := sqlite.New(dbPath)
if err != nil {
t.Fatal(err)
}
defer store.Close()
socketPath := filepath.Join(tmpDir, "test.sock")
// Set low connection limit for testing
os.Setenv("BEADS_DAEMON_MAX_CONNS", "5")
defer os.Unsetenv("BEADS_DAEMON_MAX_CONNS")
srv := NewServer(socketPath, store)
if srv.maxConns != 5 {
t.Fatalf("expected maxConns=5, got %d", srv.maxConns)
}
ctx, cancel := context.WithCancel(context.Background())
defer cancel()
go func() {
if err := srv.Start(ctx); err != nil && ctx.Err() == nil {
t.Logf("server error: %v", err)
}
}()
// Wait for server to be ready
time.Sleep(100 * time.Millisecond)
defer srv.Stop()
// Open maxConns connections and hold them
var wg sync.WaitGroup
connections := make([]net.Conn, srv.maxConns)
for i := 0; i < srv.maxConns; i++ {
conn, err := net.Dial("unix", socketPath)
if err != nil {
t.Fatalf("failed to dial connection %d: %v", i, err)
}
connections[i] = conn
// Send a long-running ping to keep connection busy
wg.Add(1)
go func(c net.Conn, idx int) {
defer wg.Done()
req := Request{
Operation: OpPing,
}
data, _ := json.Marshal(req)
c.Write(append(data, '\n'))
// Read response
reader := bufio.NewReader(c)
_, _ = reader.ReadBytes('\n')
}(conn, i)
}
// Wait for all connections to be active
time.Sleep(200 * time.Millisecond)
// Verify active connection count
activeConns := atomic.LoadInt32(&srv.activeConns)
if activeConns != int32(srv.maxConns) {
t.Errorf("expected %d active connections, got %d", srv.maxConns, activeConns)
}
// Try to open one more connection - should be rejected
extraConn, err := net.Dial("unix", socketPath)
if err != nil {
t.Fatalf("failed to dial extra connection: %v", err)
}
defer extraConn.Close()
// Send request on extra connection
req := Request{Operation: OpPing}
data, _ := json.Marshal(req)
extraConn.Write(append(data, '\n'))
// Set short read timeout to detect rejection
extraConn.SetReadDeadline(time.Now().Add(500 * time.Millisecond))
reader := bufio.NewReader(extraConn)
_, err = reader.ReadBytes('\n')
// Connection should be closed (EOF or timeout)
if err == nil {
t.Error("expected extra connection to be rejected, but got response")
}
// Close existing connections
for _, conn := range connections {
conn.Close()
}
wg.Wait()
// Wait for connection cleanup
time.Sleep(100 * time.Millisecond)
// Now should be able to connect again
newConn, err := net.Dial("unix", socketPath)
if err != nil {
t.Fatalf("failed to reconnect after cleanup: %v", err)
}
defer newConn.Close()
req = Request{Operation: OpPing}
data, _ = json.Marshal(req)
newConn.Write(append(data, '\n'))
reader = bufio.NewReader(newConn)
line, err := reader.ReadBytes('\n')
if err != nil {
t.Fatalf("failed to read response: %v", err)
}
var resp Response
if err := json.Unmarshal(line, &resp); err != nil {
t.Fatalf("failed to unmarshal response: %v", err)
}
if !resp.Success {
t.Error("expected successful ping after connection cleanup")
}
}
func TestRequestTimeout(t *testing.T) {
tmpDir := t.TempDir()
dbPath := filepath.Join(tmpDir, ".beads", "test.db")
if err := os.MkdirAll(filepath.Dir(dbPath), 0755); err != nil {
t.Fatal(err)
}
store, err := sqlite.New(dbPath)
if err != nil {
t.Fatal(err)
}
defer store.Close()
socketPath := filepath.Join(tmpDir, "test.sock")
// Set very short timeout for testing
os.Setenv("BEADS_DAEMON_REQUEST_TIMEOUT", "100ms")
defer os.Unsetenv("BEADS_DAEMON_REQUEST_TIMEOUT")
srv := NewServer(socketPath, store)
if srv.requestTimeout != 100*time.Millisecond {
t.Fatalf("expected timeout=100ms, got %v", srv.requestTimeout)
}
ctx, cancel := context.WithCancel(context.Background())
defer cancel()
go func() {
if err := srv.Start(ctx); err != nil && ctx.Err() == nil {
t.Logf("server error: %v", err)
}
}()
time.Sleep(100 * time.Millisecond)
defer srv.Stop()
conn, err := net.Dial("unix", socketPath)
if err != nil {
t.Fatalf("failed to dial: %v", err)
}
defer conn.Close()
// Send partial request and wait for timeout
conn.Write([]byte(`{"operation":"ping"`)) // Incomplete JSON
// Wait longer than timeout
time.Sleep(200 * time.Millisecond)
// Try to write - connection should be closed due to read timeout
_, err = conn.Write([]byte("}\n"))
if err == nil {
t.Error("expected connection to be closed due to timeout")
}
}
func TestMemoryPressureDetection(t *testing.T) {
tmpDir := t.TempDir()
dbPath := filepath.Join(tmpDir, ".beads", "test.db")
if err := os.MkdirAll(filepath.Dir(dbPath), 0755); err != nil {
t.Fatal(err)
}
store, err := sqlite.New(dbPath)
if err != nil {
t.Fatal(err)
}
defer store.Close()
socketPath := filepath.Join(tmpDir, "test.sock")
// Set very low memory threshold to trigger eviction
os.Setenv("BEADS_DAEMON_MEMORY_THRESHOLD_MB", "1")
defer os.Unsetenv("BEADS_DAEMON_MEMORY_THRESHOLD_MB")
srv := NewServer(socketPath, store)
// Add some entries to cache
srv.cacheMu.Lock()
for i := 0; i < 10; i++ {
path := fmt.Sprintf("/test/path/%d", i)
srv.storageCache[path] = &StorageCacheEntry{
store: store,
lastAccess: time.Now().Add(-time.Duration(i) * time.Minute),
}
}
initialSize := len(srv.storageCache)
srv.cacheMu.Unlock()
// Trigger memory pressure check (should evict entries)
srv.checkMemoryPressure()
// Check that some entries were evicted
srv.cacheMu.RLock()
finalSize := len(srv.storageCache)
srv.cacheMu.RUnlock()
if finalSize >= initialSize {
t.Errorf("expected cache eviction, but size went from %d to %d", initialSize, finalSize)
}
t.Logf("Cache evicted: %d -> %d entries", initialSize, finalSize)
}
func TestHealthResponseIncludesLimits(t *testing.T) {
tmpDir, err := os.MkdirTemp("", "bd-limits-test-*")
if err != nil {
t.Fatal(err)
}
defer os.RemoveAll(tmpDir)
dbPath := filepath.Join(tmpDir, "test.db")
socketPath := filepath.Join(tmpDir, "test.sock")
store, err := sqlite.New(dbPath)
if err != nil {
t.Fatal(err)
}
defer store.Close()
os.Setenv("BEADS_DAEMON_MAX_CONNS", "50")
defer os.Unsetenv("BEADS_DAEMON_MAX_CONNS")
srv := NewServer(socketPath, store)
ctx, cancel := context.WithCancel(context.Background())
defer cancel()
go func() {
if err := srv.Start(ctx); err != nil && ctx.Err() == nil {
t.Logf("server error: %v", err)
}
}()
time.Sleep(100 * time.Millisecond)
defer srv.Stop()
conn, err := net.Dial("unix", socketPath)
if err != nil {
t.Fatalf("failed to dial: %v", err)
}
defer conn.Close()
req := Request{Operation: OpHealth}
data, _ := json.Marshal(req)
conn.Write(append(data, '\n'))
reader := bufio.NewReader(conn)
line, err := reader.ReadBytes('\n')
if err != nil {
t.Fatalf("failed to read response: %v", err)
}
var resp Response
if err := json.Unmarshal(line, &resp); err != nil {
t.Fatalf("failed to unmarshal response: %v", err)
}
if !resp.Success {
t.Fatalf("health check failed: %s", resp.Error)
}
var health HealthResponse
if err := json.Unmarshal(resp.Data, &health); err != nil {
t.Fatalf("failed to unmarshal health response: %v", err)
}
// Verify limit fields are present
if health.MaxConns != 50 {
t.Errorf("expected MaxConns=50, got %d", health.MaxConns)
}
if health.ActiveConns < 0 {
t.Errorf("expected ActiveConns>=0, got %d", health.ActiveConns)
}
if health.MemoryAllocMB == 0 {
t.Error("expected MemoryAllocMB>0")
}
t.Logf("Health: %d/%d connections, %d MB memory", health.ActiveConns, health.MaxConns, health.MemoryAllocMB)
}

View File

@@ -150,6 +150,9 @@ type HealthResponse struct {
CacheHits int64 `json:"cache_hits"`
CacheMisses int64 `json:"cache_misses"`
DBResponseTime float64 `json:"db_response_ms"`
ActiveConns int32 `json:"active_connections"`
MaxConns int `json:"max_connections"`
MemoryAllocMB uint64 `json:"memory_alloc_mb"`
Error string `json:"error,omitempty"`
}

View File

@@ -9,6 +9,7 @@ import (
"os"
"os/signal"
"path/filepath"
"runtime"
"sort"
"strings"
"sync"
@@ -52,6 +53,12 @@ type Server struct {
startTime time.Time
cacheHits int64
cacheMisses int64
// Connection limiting
maxConns int
activeConns int32 // atomic counter
connSemaphore chan struct{}
// Request timeout
requestTimeout time.Duration
}
// NewServer creates a new RPC server
@@ -73,14 +80,32 @@ func NewServer(socketPath string, store storage.Storage) *Server {
}
}
maxConns := 100 // default
if env := os.Getenv("BEADS_DAEMON_MAX_CONNS"); env != "" {
var conns int
if _, err := fmt.Sscanf(env, "%d", &conns); err == nil && conns > 0 {
maxConns = conns
}
}
requestTimeout := 30 * time.Second // default
if env := os.Getenv("BEADS_DAEMON_REQUEST_TIMEOUT"); env != "" {
if timeout, err := time.ParseDuration(env); err == nil && timeout > 0 {
requestTimeout = timeout
}
}
return &Server{
socketPath: socketPath,
storage: store,
storageCache: make(map[string]*StorageCacheEntry),
maxCacheSize: maxCacheSize,
cacheTTL: cacheTTL,
shutdownChan: make(chan struct{}),
startTime: time.Now(),
socketPath: socketPath,
storage: store,
storageCache: make(map[string]*StorageCacheEntry),
maxCacheSize: maxCacheSize,
cacheTTL: cacheTTL,
shutdownChan: make(chan struct{}),
startTime: time.Now(),
maxConns: maxConns,
connSemaphore: make(chan struct{}, maxConns),
requestTimeout: requestTimeout,
}
}
@@ -131,7 +156,20 @@ func (s *Server) Start(ctx context.Context) error {
return fmt.Errorf("failed to accept connection: %w", err)
}
go s.handleConnection(conn)
// Try to acquire connection slot (non-blocking)
select {
case s.connSemaphore <- struct{}{}:
// Acquired slot, handle connection
go func(c net.Conn) {
defer func() { <-s.connSemaphore }() // Release slot
atomic.AddInt32(&s.activeConns, 1)
defer atomic.AddInt32(&s.activeConns, -1)
s.handleConnection(c)
}(conn)
default:
// Max connections reached, reject immediately
conn.Close()
}
}
}
@@ -218,7 +256,7 @@ func (s *Server) handleSignals() {
s.Stop()
}
// runCleanupLoop periodically evicts stale storage connections
// runCleanupLoop periodically evicts stale storage connections and checks memory pressure
func (s *Server) runCleanupLoop() {
s.cleanupTicker = time.NewTicker(5 * time.Minute)
defer s.cleanupTicker.Stop()
@@ -226,6 +264,7 @@ func (s *Server) runCleanupLoop() {
for {
select {
case <-s.cleanupTicker.C:
s.checkMemoryPressure()
s.evictStaleStorage()
case <-s.shutdownChan:
return
@@ -233,6 +272,71 @@ func (s *Server) runCleanupLoop() {
}
}
// checkMemoryPressure monitors memory usage and triggers aggressive eviction if needed
func (s *Server) checkMemoryPressure() {
var m runtime.MemStats
runtime.ReadMemStats(&m)
// Memory thresholds (configurable via env var)
const defaultThresholdMB = 500
thresholdMB := defaultThresholdMB
if env := os.Getenv("BEADS_DAEMON_MEMORY_THRESHOLD_MB"); env != "" {
var threshold int
if _, err := fmt.Sscanf(env, "%d", &threshold); err == nil && threshold > 0 {
thresholdMB = threshold
}
}
allocMB := m.Alloc / 1024 / 1024
if allocMB > uint64(thresholdMB) {
fmt.Fprintf(os.Stderr, "Warning: High memory usage detected (%d MB), triggering aggressive cache eviction\n", allocMB)
s.aggressiveEviction()
runtime.GC() // Suggest garbage collection
}
}
// aggressiveEviction evicts 50% of cached storage to reduce memory pressure
func (s *Server) aggressiveEviction() {
toClose := []storage.Storage{}
s.cacheMu.Lock()
if len(s.storageCache) == 0 {
s.cacheMu.Unlock()
return
}
// Build sorted list by last access
type cacheItem struct {
path string
entry *StorageCacheEntry
}
items := make([]cacheItem, 0, len(s.storageCache))
for path, entry := range s.storageCache {
items = append(items, cacheItem{path, entry})
}
sort.Slice(items, func(i, j int) bool {
return items[i].entry.lastAccess.Before(items[j].entry.lastAccess)
})
// Evict oldest 50%
numToEvict := len(items) / 2
for i := 0; i < numToEvict; i++ {
toClose = append(toClose, items[i].entry.store)
delete(s.storageCache, items[i].path)
}
s.cacheMu.Unlock()
// Close without holding lock
for _, store := range toClose {
if err := store.Close(); err != nil {
fmt.Fprintf(os.Stderr, "Warning: failed to close evicted storage: %v\n", err)
}
}
}
// evictStaleStorage removes idle connections and enforces cache size limits
func (s *Server) evictStaleStorage() {
now := time.Now()
@@ -291,6 +395,11 @@ func (s *Server) handleConnection(conn net.Conn) {
writer := bufio.NewWriter(conn)
for {
// Set read deadline for the next request
if err := conn.SetReadDeadline(time.Now().Add(s.requestTimeout)); err != nil {
return
}
line, err := reader.ReadBytes('\n')
if err != nil {
return
@@ -306,6 +415,11 @@ func (s *Server) handleConnection(conn net.Conn) {
continue
}
// Set write deadline for the response
if err := conn.SetWriteDeadline(time.Now().Add(s.requestTimeout)); err != nil {
return
}
resp := s.handleRequest(&req)
s.writeResponse(writer, resp)
}
@@ -488,6 +602,10 @@ func (s *Server) handlePing(_ *Request) Response {
func (s *Server) handleHealth(req *Request) Response {
start := time.Now()
// Get memory stats for health response
var m runtime.MemStats
runtime.ReadMemStats(&m)
store, err := s.getStorageForRequest(req)
if err != nil {
data, _ := json.Marshal(HealthResponse{
@@ -541,6 +659,9 @@ func (s *Server) handleHealth(req *Request) Response {
CacheHits: atomic.LoadInt64(&s.cacheHits),
CacheMisses: atomic.LoadInt64(&s.cacheMisses),
DBResponseTime: dbResponseMs,
ActiveConns: atomic.LoadInt32(&s.activeConns),
MaxConns: s.maxConns,
MemoryAllocMB: m.Alloc / 1024 / 1024,
}
if dbError != "" {