Load Balancer Tab: HAProxy Statistics
How HAProxy shares out your traffic
Watch your load balancer
On a server that runs HAProxy, the Load Balancer tab shows how traffic is shared out: which backend servers are up, how fast they answer, which status codes they return, and the latest requests. It is read-only; nothing here changes your HAProxy setup.
01 · Summary
The top card shows the HAProxy version and these numbers. Hover the info icon of any of them for a short explanation.
Time since HAProxy started. Every counter on the tab counts from that moment.
Open client connections, against the connection limit.
HTTP requests handled since start.
New connections per second, now and at the peak.
Secure connections set up per second, now and at the peak.
Bytes sent since start, and the current rate.
How much of the time HAProxy has nothing to do. Low means HAProxy itself is the bottleneck.
Times the connection limit was reached, and log lines lost. Both should be 0.
The tab refreshes itself every 10 seconds while Every 10s is ticked. Refresh reloads it at once.
02 · Backends and servers
One block per backend: its name, status, balancing method, requests, active sessions and queue. Under it, one row per server.
| Column | What it tells you |
|---|---|
| Status | UP, DOWN, or MAINT / DRAIN when a server is taken out on purpose. |
| Health check | The last check result and how long it took, plus failed checks. |
| Share | The part of the backend's traffic this server took. |
| Avg connect / response / total | Average times, over each server's last 1024 requests. |
| 2xx / 3xx / 4xx / 5xx | Responses by status class; 5xx turns red. |
| Errors conn / resp | Failed connections and failed responses; red above 0. |
| State for | Time since the server last changed state, and its total downtime. |
The table also has sessions now and peak, requests, maximum times, retries and redispatches, and traffic in and out.
03 · Frontends
One row per entry point: status, sessions against their limit, session and request rates, requests, responses by status class, denied and bad requests, TLS sessions with the share reused, and traffic in and out.
04 · Request log
- The newest requests come first. Choose Last 100, 300, 1000 or 2000 lines.
- Type in Filter to keep only lines with a server name, status, IP address or URL.
- Server errors and failed connections are red, client errors are amber.
- The timers in each line are request / queue / connect / response / total, in milliseconds, then the status, the bytes and the backend and server that answered.
HAProxy also gets its own card on the Services tab, to restart or reload it, and its own tab in the Logs bar. To change its configuration, use the Terminal.