Our agents are deployed to Livekit Cloud, and over the past week we’re seeing intermittent errors from the Python Agents OpenTelemetry log exporter when exporting logs.
Failed to export logs batch code: 429, reason:
QuotaStatusNetworkError
type.googleapis.com/google.rpc.BadRequest
The failing request is via processor thread OtelBatchLogRecordProcessor to:
We’ve confirmed this does not appear to impact call/session behavior operationally, but wanted to check whether this indicates anything we should tune on our side, or if it points to a LiveKit Cloud observability ingestion quota/rate-limit that the LiveKit team should review.
@D3x_Ops, the 429 with QuotaStatusNetworkError against production.livekit.cloud/observability/logs/otlp/v0 is a Cloud-side ingestion throttle, not a client misconfiguration. You’re right that drop-rate from a BatchLogRecordProcessor doesn’t affect call behavior; the exporter retries with backoff and only drops on queue overflow.
Two practical levers on your side:
Lower log volume. Drop verbose loggers via logging.getLogger("livekit.agents").setLevel(logging.INFO) (or WARNING). Also silence chatty deps like aiohttp, httpx.
Tune the batch processor via standard OTel env vars: bump OTEL_BLRP_SCHEDULE_DELAY=15000 and OTEL_BLRP_MAX_EXPORT_BATCH_SIZE=1024 to send fewer, larger batches.
For the actual quota number and whether it can be raised on your project, that’s a question for @CWilson or @darryncampbell. They have visibility into the per-project observability ingestion limits that aren’t documented publicly.