Statistics

Monitor per-origin traffic, latency, failures, and failover behavior for your Load Balancers.

Every Load Balancer records detailed statistics about the traffic it routes, broken down per origin and per origin group. Use them to verify your traffic distribution, spot a slow or failing origin, and understand how often failover happens.

Statistics are available in the dashboard and through the API. See the API reference for the endpoint details.

Summary metrics

Metric	Description
Total requests	Requests handled by the Load Balancer. Requests served from the CDN cache are not included.
Total traffic	Bytes served from your origins through the Load Balancer.
Average origin RTT	Representative origin response time (time to first byte) across all origins.
Requests with failover	How many requests needed more than one origin before succeeding, and the average number of extra origins tried.
All origins exhausted	Requests that failed on every origin and returned 502 to the client.
Sticky session hits	Requests routed by an existing sticky session instead of the balancing method.
Requests by method	Request counts per HTTP method.

Per-origin time series

• Requests by origin and by origin group: how traffic was distributed. Use this to confirm weights and priorities behave as intended.
• Traffic by origin: bytes served per origin.
• Origin latency: response time per origin, as mean and 50th, 90th, 95th, and 99th percentiles.
• Origin failures: failed attempts per origin, broken down by status code. Connection-level failures (timeouts, refused connections) are reported as 502 or 504.

Interpreting the numbers

Requests are attributed to the origin that produced the served response. Attempts that failed and were failed over to another origin are not counted as requests for the failing origin; they appear in that origin’s failure counts instead. If you compare bunny statistics with your own origin access logs, the difference corresponds to the attempts that failed over.

Failures count attempts, not client errors. A single client request that fails over across three origins records a failure on the first two and a request on the third, while the client sees one successful response.

Time ranges

Statistics can be grouped daily or hourly. A single query can cover up to 90 days. The account-level overview shows month-to-date totals for every Load Balancer, including how many Pull Zones each one is attached to.

Billing usage

The usage endpoint reports month-to-date billable requests against the included allowance for each Load Balancer, along with when the allowance resets. See Pricing for how requests are billed.