Skip to content

Traffic and Reachability ​

Traffic Stats ​

frpc doesn't count traffic per proxy, so Podux reads it from the frps dashboard (webServer). First enable the dashboard on frps:

toml
# frps.toml
webServer.addr = "0.0.0.0"
webServer.port = 7500
webServer.user = "admin"
webServer.password = "a strong password"

Then, in the Podux server settings, enter the frps dashboard address (e.g. http://frps.example.com:7500), username and password, and click Test connection. You then get:

  • The "Reachability / traffic" column of the proxy list and server detail: each proxy's traffic in and out today and current connections; click for a 7-day chart
  • "frps status" on the server detail page: frps version, total traffic today, connections and clients online
  • "Traffic today" on the dashboard: the sum over servers with a dashboard

frps keeps the traffic counts in memory, so they reset when frps restarts. Podux queries each frps at most once every 15 seconds.

WARNING

The dashboard password is stored in Podux's database. If the dashboard is reachable from the internet, use a strong password or only allow the Podux host.

External Reachability Probes ​

An "online" proxy in frpc only means the tunnel is up; it doesn't guarantee it can be reached from outside (a firewall on the frps host may block the port, or the domain may not resolve). With a target cycle of 1 minute, from the Podux host, Podux probes the online proxies of running servers:

Proxy typeProbe
TCPConnect to server address:remote port
HTTPRequest frps's HTTP port with the proxy's domain and first location. frps's own 404 page (no matching proxy) or 502/503/504 count as failures; any other status from the service, including 401 and 404, counts as reachable
HTTPSTLS handshake with frps's HTTPS port using the proxy's domain

UDP, STCP, SUDP, XTCP and TCPMUX proxies are not probed, nor are proxies with remote port 0 (chosen by frps) or only wildcard domains.

TCP, HTTP and HTTPS proxy forms have an External reachability probe switch under Network Config. It is on by default, including for existing proxies without this setting, and can be turned off per proxy. Turning it off only stops Podux's independent external probes; it does not disable the proxy, change tunnel settings or affect frpc's local service health checks.

When off, the "Reachability / traffic" column in the proxy list and server detail shows a gray "Probe disabled" badge instead of old reachable or unreachable results. With an frps dashboard configured, traffic stats, current connections and click-to-view 7-day traffic history remain available. Editing and cloning preserve the setting; switching to an unsupported proxy type hides the switch without overwriting its value.

HTTP/HTTPS probes use frps's vhost ports and subdomainHost: read from frps when a dashboard is configured, otherwise 80 and 443, and subdomain-only proxies are skipped.

If frps sits behind an HTTPS reverse proxy (for example nginx terminating TLS on 443) and its vhost ports are not reachable from outside, turn on Probe through public HTTPS in the server's frps dashboard section. HTTP and HTTPS proxies are then probed on their domain's port 443 the way users reach them; the certificate is not checked. Subdomain proxies still need the dashboard for subdomainHost.

Within each round, servers are probed in parallel, with probe starts staggered and rate-limited separately for each server record: at most 4 concurrent probes and at least max(250ms, 1min / N) between starts (N is that server's target count for the current round). This spreads short-lived connection bursts. The next round waits for all servers to finish, so many or slow targets can delay its start; rounds never overlap and missed rounds are not replayed to catch up. This only reduces bursts and does not guarantee elimination of FRP connection-pool-full errors.

Each completed probe immediately updates its reachability result, shown in the "Reachability / traffic" column; alerts are evaluated after each server's complete round. With the "Proxy unreachable from outside" alert on, a proxy failing probes in 3 consecutive complete rounds triggers an alert, and another when it recovers. This usually takes about 3 minutes, but depends on round completion and takes longer with many or slow targets.

Stopping the service cancels queued and in-flight probes and waits for them to exit; these cancellations are not recorded as failures and do not cause false unreachable alerts.

TIP

Probes go out from the Podux host. If it shares a LAN with frps, or your router doesn't support reaching its own public address (hairpin NAT), results can differ from real outside access.

Released under the Apache 2.0 License.