perf: expose and reduce message publish latency

This commit is contained in:
richjohnrambo
2026-09-27 21:08:35 +08:00
parent 8e778726e0
commit bde02e8eac
15 changed files with 274 additions and 98 deletions
+47 -18
View File
@@ -17,22 +17,25 @@ type observation struct {
}
type deliveryRecorder struct {
machine string
localUsers int
latency *DurationHistogram
disconnects atomic.Int64
received atomic.Int64
runID atomic.Value
historyMode atomic.Bool
machine string
localUsers int
latency *DurationHistogram
localLatency *DurationHistogram
crossLatency *DurationHistogram
disconnects atomic.Int64
received atomic.Int64
runID atomic.Value
historyMode atomic.Bool
mu sync.Mutex
observed map[string]*observation
lastSeq []map[string]int64
duplicates int64
outOfOrder int64
ordering []OrderingViolation
malformed int64
history map[string]bool
mu sync.Mutex
observed map[string]*observation
lastSeq []map[string]int64
duplicates int64
outOfOrder int64
ordering []OrderingViolation
malformed int64
crossClockSkewSamples int64
history map[string]bool
}
func newDeliveryRecorder(runID, machine string, localUsers int) *deliveryRecorder {
@@ -42,6 +45,7 @@ func newDeliveryRecorder(runID, machine string, localUsers int) *deliveryRecorde
}
recorder := &deliveryRecorder{
machine: machine, localUsers: localUsers, latency: NewDurationHistogram(),
localLatency: NewDurationHistogram(), crossLatency: NewDurationHistogram(),
observed: make(map[string]*observation), lastSeq: lastSeq, history: make(map[string]bool),
}
recorder.runID.Store(runID)
@@ -57,8 +61,9 @@ func (r *deliveryRecorder) onData(account AccountRecord, data dataPacket) {
return
}
var content struct {
RunID string `json:"run_id"`
SentAt int64 `json:"sent_at_unix_nano"`
RunID string `json:"run_id"`
SenderMachine string `json:"sender_machine"`
SentAt int64 `json:"sent_at_unix_nano"`
}
if json.Unmarshal(data.Content, &content) != nil || content.RunID != runID {
r.mu.Lock()
@@ -108,10 +113,31 @@ func (r *deliveryRecorder) onData(account AccountRecord, data dataPacket) {
r.lastSeq[index][data.Topic] = data.Seq
}
if content.SentAt > 0 {
r.latency.Observe(time.Since(time.Unix(0, content.SentAt)))
delay := time.Since(time.Unix(0, content.SentAt))
r.latency.Observe(delay)
senderMachine := content.SenderMachine
if senderMachine == "" {
senderMachine = senderMachineFromClientMid(runID, mid)
}
if senderMachine == r.machine {
r.localLatency.Observe(delay)
} else if delay < 0 {
r.crossClockSkewSamples++
} else {
r.crossLatency.Observe(delay)
}
}
}
func senderMachineFromClientMid(runID, mid string) string {
prefix := strings.TrimPrefix(mid, runID+"-")
parts := strings.SplitN(prefix, "-", 3)
if len(parts) < 2 {
return ""
}
return parts[1]
}
func (r *deliveryRecorder) metricCounts() (received, duplicates, outOfOrder, disconnects int64) {
r.mu.Lock()
duplicates = r.duplicates
@@ -134,6 +160,8 @@ func (r *deliveryRecorder) onDisconnect(_ AccountRecord, _ error) {
func (r *deliveryRecorder) summarize(base AgentSummary, ledger []AcceptedMessage) AgentSummary {
base.DeliveryLatency = r.latency.Snapshot()
base.DeliveryLatencyLocal = r.localLatency.Snapshot()
base.DeliveryLatencyCross = r.crossLatency.Snapshot()
base.Disconnects = r.disconnects.Load()
runID, _ := r.runID.Load().(string)
accepted := make(map[string]AcceptedMessage, len(ledger))
@@ -143,6 +171,7 @@ func (r *deliveryRecorder) summarize(base AgentSummary, ledger []AcceptedMessage
defer r.mu.Unlock()
base.Duplicates = r.duplicates
base.OutOfOrder = r.outOfOrder
base.CrossClockSkewSamples = r.crossClockSkewSamples
base.OrderingViolations = append([]OrderingViolation(nil), r.ordering...)
if r.malformed > 0 {
base.Errors = append(base.Errors, fmt.Sprintf("malformed run messages=%d", r.malformed))