Skip to main content

OTEL-025: batch timeout below 100ms

Severity: warn (advisory)

Rule Details​

timeout is the maximum time a batch can wait before being flushed. Anything under ~100ms means most batches flush on the timeout rather than on send_batch_size, so you pay the per-batch overhead (gRPC headers, compression setup, TLS handshakes for keepalive-less exporters) on tiny batches. If you need low end-to-end latency, set other knobs; do not starve the batcher.

This rule fires when batch.timeout < 100ms.

Options​

FieldConstraint
timeoutShould be ≥ 100ms

Examples​

Avoid
processors:
  batch:
    timeout: 10ms                    # below 100ms
    send_batch_size: 1024
Prefer
processors:
  batch:
    timeout: 1s
    send_batch_size: 1024

When Not To Use It​

Ultra-low-latency tracing pipelines where you deliberately trade throughput for freshness and have measured that a sub-100ms timeout is actually needed.

  • OTEL-023 — batch send_batch_max_size unset
  • OTEL-024 — send_batch_max_size < send_batch_size
  • OTEL-026 — batch timeout above 60s

Version​

Available since augur v0.1.0.

Further Reading​

Resources​