Reporter

Reporter is responsible for batching and transmitting collected signals to the Sentinel backend (POST /sen/ss). It supports both time-based and size-based flush triggers, with automatic retry on failure.

Responsibilities

Responsibility	Detail
Batch flush	Sends accumulated signals when a threshold is met
Time-based interval	Periodic forced flush at signalThresholdMs intervals
Size-based trigger	Immediate flush when unsent signal count exceeds signalThresholdSize
Token rotation	Stores any new token returned by the server
Config propagation	Forwards any new config from the server to Sentinel.updateConfig()
Retry with back-off	Uses RetryQueue for resilient delivery

Flush Triggers

Trigger	Condition	Force?
Signal added	unsentSignals.length >= sizeThreshold	No (respects threshold)
Interval timer	Every timeThresholdMs milliseconds	Yes (flushes any pending signals)
App foreground	AppStateCollector detects tab activation	Yes

Public API

Method	Signature	Description
start	() => void	Begin periodic flushing (requires thresholds to be set)
stop	() => void	Stop periodic flushing
flushIfNeeded	(force?: boolean) => Promise<void>	Flush signals if threshold met, or immediately if force = true
setThresholds	(thresholds: ReporterThresholds) => void	Update time/size thresholds and reschedule interval

Flush Payload

interface ReporterRequest {
  data: {
    token: string;
    signals: Array<Signal>;
  };
}

Signals are sliced to sizeThreshold per flush call. The request is sent to POST /sen/ss on the fsp domain.

Flush Response

interface ReporterResponse {
  data: {
    token: string;
    config: SentinelConfig;
  };
}

On success:

1. New token is stored via TokenManager.
2. New config is applied via Sentinel.updateConfig() → ConfigManager.
3. Sent signal IDs are removed from SignalStore.

Concurrency Control

The reporter uses a isFlushing flag to prevent overlapping flush calls. If flushIfNeeded is called while a flush is in progress, a pendingFlush flag is set. Once the current flush completes, the pending flush is re-triggered.

Retry Strategy

Each flush is wrapped in a RetryQueue:

Parameter	Value
Max attempts	3
Base delay	200 ms
Back-off	Exponential (200ms → 400ms → 800ms)

If all retries fail, sen.reporter_flush_failed is logged and the signals remain in the store for the next attempt.

Threshold Lifecycle

Thresholds are initially null. start() is a no-op until setThresholds() has been called. This prevents premature flushing before the server config arrives.

When thresholds are updated while the reporter is running, the interval timer is rescheduled with the new timeThresholdMs.

---

Last updated: 2026-02-24