fix: adding monitoring
This commit is contained in:
16
internal/format.go
Normal file
16
internal/format.go
Normal file
@@ -0,0 +1,16 @@
|
||||
package internal
|
||||
|
||||
import "fmt"
|
||||
|
||||
func FormatBytes(b uint64) string {
|
||||
const unit = 1024
|
||||
if b < unit {
|
||||
return fmt.Sprintf("%d B", b)
|
||||
}
|
||||
div, exp := uint64(unit), 0
|
||||
for n := b / unit; n >= unit; n /= unit {
|
||||
div *= unit
|
||||
exp++
|
||||
}
|
||||
return fmt.Sprintf("%.2f %cB", float64(b)/float64(div), "KMGTPE"[exp])
|
||||
}
|
||||
102
internal/hub.go
102
internal/hub.go
@@ -1,6 +1,7 @@
|
||||
package internal
|
||||
|
||||
import (
|
||||
"context"
|
||||
"fmt"
|
||||
"log"
|
||||
"sync"
|
||||
@@ -17,6 +18,11 @@ const (
|
||||
maxMessageSize = 512
|
||||
)
|
||||
|
||||
const (
|
||||
minHighWaterMarkMapRebuild = 100
|
||||
mapRebuildThreshold = 4
|
||||
)
|
||||
|
||||
type Client struct {
|
||||
ID string
|
||||
Conn *websocket.Conn
|
||||
@@ -36,34 +42,86 @@ func NewClient(conn *websocket.Conn, subscribedPath string) *Client {
|
||||
}
|
||||
|
||||
type Hub struct {
|
||||
Clients map[*Client]bool
|
||||
Broadcast chan []byte
|
||||
Register chan *Client
|
||||
Unregister chan *Client
|
||||
path string
|
||||
maxClients int
|
||||
Clients map[*Client]bool
|
||||
Broadcast chan []byte
|
||||
Register chan *Client
|
||||
Unregister chan *Client
|
||||
monitor *MemoryMonitor
|
||||
highWaterMark int
|
||||
}
|
||||
|
||||
func NewHub() *Hub {
|
||||
func NewHub(path string, maxClients int) *Hub {
|
||||
log.Printf("[%s] Hub created with max clients: %d", path, maxClients)
|
||||
return &Hub{
|
||||
path: path,
|
||||
maxClients: maxClients,
|
||||
monitor: NewMemoryMonitor(),
|
||||
Broadcast: make(chan []byte, 256),
|
||||
Register: make(chan *Client, 10),
|
||||
Unregister: make(chan *Client, 10),
|
||||
Register: make(chan *Client, maxClients),
|
||||
Unregister: make(chan *Client, maxClients),
|
||||
Clients: make(map[*Client]bool),
|
||||
}
|
||||
}
|
||||
|
||||
func (h *Hub) Run() {
|
||||
// Run starts the hub event loop. It exits when ctx is cancelled.
|
||||
func (h *Hub) Run(ctx context.Context) {
|
||||
monitorTicker := time.NewTicker(MonitorInterval)
|
||||
go func() {
|
||||
defer func() {
|
||||
monitorTicker.Stop()
|
||||
// On shutdown, close every client's Send channel so WritePump sends
|
||||
// a WebSocket close frame and exits. Also expire the read deadline so
|
||||
// blocked ReadMessage() calls in ReadPump return immediately instead
|
||||
// of waiting up to pongWait (60 s).
|
||||
for client := range h.Clients {
|
||||
close(client.Send)
|
||||
client.Conn.SetReadDeadline(time.Now())
|
||||
}
|
||||
log.Printf("[%s] Hub stopped\n", h.path)
|
||||
}()
|
||||
|
||||
for {
|
||||
select {
|
||||
case <-ctx.Done():
|
||||
log.Printf("[%s] Hub shutting down\n", h.path)
|
||||
return
|
||||
|
||||
case client := <-h.Register:
|
||||
if len(h.Clients) >= h.maxClients {
|
||||
close(client.Send)
|
||||
client.Conn.Close()
|
||||
log.Printf("[%s] Rejected client %s (max %d reached)\n", h.path, client.ID, h.maxClients)
|
||||
break
|
||||
}
|
||||
h.Clients[client] = true
|
||||
log.Printf("Client registered %s\n", client.ID)
|
||||
if len(h.Clients) > h.highWaterMark {
|
||||
h.highWaterMark = len(h.Clients)
|
||||
}
|
||||
log.Printf("[%s] Client registered %s\n", h.path, client.ID)
|
||||
|
||||
case client := <-h.Unregister:
|
||||
if _, ok := h.Clients[client]; ok {
|
||||
delete(h.Clients, client)
|
||||
close(client.Send)
|
||||
|
||||
// Rebuild the map when the live set has dropped to less than
|
||||
// 1/mapRebuildThreshold of the peak, so the old backing buckets
|
||||
// are released to the GC.
|
||||
if h.highWaterMark >= minHighWaterMarkMapRebuild &&
|
||||
len(h.Clients) < h.highWaterMark/mapRebuildThreshold {
|
||||
rebuilt := make(map[*Client]bool, len(h.Clients))
|
||||
for c := range h.Clients {
|
||||
rebuilt[c] = true
|
||||
}
|
||||
h.Clients = rebuilt
|
||||
h.highWaterMark = len(h.Clients)
|
||||
log.Printf("[%s] Clients map rebuilt: %d active clients\n", h.path, len(h.Clients))
|
||||
}
|
||||
log.Printf("[%s] Client Unregistered %s\n", h.path, client.ID)
|
||||
}
|
||||
log.Printf("Client Unregistered %s\n", client.ID)
|
||||
|
||||
case message := <-h.Broadcast:
|
||||
for client := range h.Clients {
|
||||
select {
|
||||
@@ -71,9 +129,18 @@ func (h *Hub) Run() {
|
||||
default:
|
||||
close(client.Send)
|
||||
delete(h.Clients, client)
|
||||
log.Printf("Client %s removed (slow/disconnected)", client.ID)
|
||||
log.Printf("[%s] Client %s removed (slow/disconnected)\n", h.path, client.ID)
|
||||
}
|
||||
}
|
||||
|
||||
case <-monitorTicker.C:
|
||||
current, peak := h.monitor.Snapshot()
|
||||
clientLength := len(h.Clients)
|
||||
if clientLength > 0 {
|
||||
log.Printf("[%s] connected clients: %d | heap alloc: %s | peak heap alloc: %s",
|
||||
h.path, clientLength, FormatBytes(current), FormatBytes(peak),
|
||||
)
|
||||
}
|
||||
}
|
||||
}
|
||||
}()
|
||||
@@ -82,7 +149,10 @@ func (h *Hub) Run() {
|
||||
func WritePump(c *Client, h *Hub) {
|
||||
pingTicker := time.NewTicker(pingPeriod)
|
||||
defer func() {
|
||||
h.Unregister <- c
|
||||
select {
|
||||
case h.Unregister <- c:
|
||||
default:
|
||||
}
|
||||
pingTicker.Stop()
|
||||
c.Conn.Close()
|
||||
}()
|
||||
@@ -103,7 +173,7 @@ func WritePump(c *Client, h *Hub) {
|
||||
}
|
||||
w.Write(message)
|
||||
|
||||
// Queue queued messages in the same buffer (optional optimization)
|
||||
// Flush any messages that queued up while we were writing.
|
||||
n := len(c.Send)
|
||||
for i := 0; i < n; i++ {
|
||||
w.Write(<-c.Send)
|
||||
@@ -120,13 +190,15 @@ func WritePump(c *Client, h *Hub) {
|
||||
return
|
||||
}
|
||||
}
|
||||
|
||||
}
|
||||
}
|
||||
|
||||
func ReadPump(c *Client, h *Hub) {
|
||||
defer func() {
|
||||
h.Unregister <- c
|
||||
select {
|
||||
case h.Unregister <- c:
|
||||
default:
|
||||
}
|
||||
c.Conn.Close()
|
||||
}()
|
||||
|
||||
|
||||
60
internal/memory_monitor.go
Normal file
60
internal/memory_monitor.go
Normal file
@@ -0,0 +1,60 @@
|
||||
package internal
|
||||
|
||||
import (
|
||||
"context"
|
||||
"log"
|
||||
"runtime"
|
||||
"sync/atomic"
|
||||
"time"
|
||||
)
|
||||
|
||||
const MonitorInterval = 30 * time.Second
|
||||
|
||||
type MemoryMonitor struct {
|
||||
peakAlloc atomic.Uint64
|
||||
}
|
||||
|
||||
func NewMemoryMonitor() *MemoryMonitor {
|
||||
return &MemoryMonitor{}
|
||||
}
|
||||
|
||||
// Snapshot reads current heap allocation, updates the peak, and returns both.
|
||||
func (m *MemoryMonitor) Snapshot() (currentAlloc, peakAlloc uint64) {
|
||||
var ms runtime.MemStats
|
||||
runtime.ReadMemStats(&ms)
|
||||
currentAlloc = ms.HeapAlloc
|
||||
for {
|
||||
peak := m.peakAlloc.Load()
|
||||
if currentAlloc <= peak {
|
||||
return currentAlloc, peak
|
||||
}
|
||||
if m.peakAlloc.CompareAndSwap(peak, currentAlloc) {
|
||||
return currentAlloc, currentAlloc
|
||||
}
|
||||
}
|
||||
}
|
||||
|
||||
// Run starts a periodic monitor loop that calls logFn with each snapshot.
|
||||
// It blocks until ctx is cancelled.
|
||||
func (m *MemoryMonitor) Run(ctx context.Context, logFn func(currentAlloc, peakAlloc uint64)) {
|
||||
ticker := time.NewTicker(MonitorInterval)
|
||||
defer ticker.Stop()
|
||||
|
||||
for {
|
||||
select {
|
||||
case <-ctx.Done():
|
||||
return
|
||||
case <-ticker.C:
|
||||
current, peak := m.Snapshot()
|
||||
logFn(current, peak)
|
||||
}
|
||||
}
|
||||
}
|
||||
|
||||
// DefaultLogFn returns a log function with the given prefix.
|
||||
func DefaultLogFn(prefix string) func(currentAlloc, peakAlloc uint64) {
|
||||
return func(currentAlloc, peakAlloc uint64) {
|
||||
log.Printf("[%s] heap alloc: %s | peak heap alloc: %s",
|
||||
prefix, FormatBytes(currentAlloc), FormatBytes(peakAlloc))
|
||||
}
|
||||
}
|
||||
Reference in New Issue
Block a user