Ana içeriğe geç

Analytics Tab

The API Analysis tab provides summary information about the API Proxy's performance, endpoint usage intensity, and client activity.

By selecting the refresh interval from the Choose your Refresh Time input, data is automatically refreshed until the page is changed.

The visual containing the Analytics tab is shown below:

Analytics Tab

Routing Diagnostics

The Analytics tab includes a Routing Diagnostics section that summarizes routing problems seen on requests this API Proxy sends to the backend. This section shows, in aggregate, the delays and failures that occur while connecting, reading, and leasing a connection from the connection pool.

Info

This data is collected automatically whenever the Performance Metrics option is enabled in the connector's Log Settings (the default state); it is also collected even when that option is disabled, as long as Tracing is active for the request or the request ends in a failure/block. Data is only produced for requests that go to the backend; requests served from cache or that never reach the backend are not included here.

Routing Diagnostics charts on the Analytics tab

Failure Reason Distribution

A bar chart showing the distribution of routing failures by reason over the selected time range. Clicking a bar filters the records for that failure reason in the API Traffic table.

Possible failure reasons:

Failure ReasonDescription
Pool TimeoutCould not lease a connection from the pool within the configured time — the pool may be exhausted
Connect TimeoutTCP connection to the backend could not be established — backend down, wrong host/port, or network/firewall blocking
DNS FailureBackend host name could not be resolved
TLS Handshake FailureTLS handshake with the backend failed — certificate, protocol, or cipher mismatch
Read Timeout (no response)No first byte was received before the read timeout on a reused connection — possibly a stale/broken connection
Read Timeout (slow backend)First byte was received but the backend did not finish the response in time
Backend ClosedThe backend closed the connection unexpectedly
Client ClosedThe client aborted the connection before the response completed
No Healthy UpstreamNo healthy backend address was available to route to
Circuit OpenThe circuit breaker was open for this route; the request was not sent to the backend
Retries ExhaustedAll retry/failover attempts failed
Upstream HTTP ErrorThe backend received the request and returned an HTTP error status
UnknownThe failure could not be classified from the available signals

Phase Latency (p50/p95/p99)

A chart showing how much time requests to the backend spend in each phase. p50/p95/p99 latency values (in milliseconds) are calculated for the following phases:

PhaseDescription
SelectionChoosing the target address via load balancing and circuit-breaker checks
DNSResolving the backend host name
TCP ConnectEstablishing the TCP connection to the backend
TLS HandshakeTLS handshake for HTTPS connections
Time to First Byte (TTFB)Time from sending the request to receiving the backend's first response byte
Body ReadReading the entire response body
Pool WaitTime spent waiting to lease a suitable connection from the connection pool

This breakdown helps you tell whether slowness is happening during connection setup (DNS/TCP/TLS), on the backend's response (TTFB/body read), or in the connection pool (pool wait).

Other Summaries

  • Top Failing Proxies: The API Proxies producing the most failures among requests with a routing failure record
  • Slow Client vs Slow Backend (average): An average comparison of the time spent writing the response to the client versus the backend processing time — indicates whether the slowness is on the client/network side or the backend side
  • Top Upstreams: The upstream addresses (IP:Port) with the most failures
Tip

For the Routing Diagnostics details of a single request (failure reason, confidence level, phase timing, recommended action, etc.), see the Detailed View window on the API Traffic tab.