EvergenEnergy / sparkplug-host / 29222541534

85%

Build Type

push

github

Committed by web-flow

Commit Message

fix: set MQTT write timeout so AutoReconnect recovers from half-open connections (#48)

* fix: set MQTT write timeout so AutoReconnect recovers from half-open connections

paho's AutoReconnect only fires after the client observes ConnectionLost. On
a half-open TCP connection (e.g. broker recreated behind a proxy/sidecar with
no RST reaching the client) the keepalive PING is a socket write, and with
paho's default of no write timeout that write blocks forever on the dead
socket. ConnectionLost is never reported, AutoReconnect never fires, and the
host application sits silently disconnected.

Set a default MQTT write timeout (10s) and an explicit keepalive (30s, paho's
default) in initClients() so a stalled PING fails and the lost connection is
surfaced for AutoReconnect to recover. Both are configurable via the new
WithMqttWriteTimeout and WithMqttKeepAlive options; WithMqttWriteTimeout(0)
restores the previous blocking behaviour.

* feat: add WithMqttPingTimeout and correct write-timeout mechanism docs

Addresses review feedback on the write-timeout mechanism:

- Correct the doc comments: in paho v1.5.1 WriteTimeout is applied only to
  outbound application writes in the outgoing worker (net.go SetWriteDeadline),
  NOT to the keepalive PING (ping.go writes it with no deadline). The real
  benefit is unsticking a blocked outbound write on a half-open connection whose
  send buffer has filled (which can also stall the keepalive goroutine and stop
  the ping-timeout check from running). Reworded options.go and host.go
  accordingly.
- Add WithMqttPingTimeout: for an idle half-open connection the small PINGREQ is
  written without blocking and detection is governed by the ping timeout, so
  expose it as the knob for idle-detection latency. Default 10s (paho default).
- Document the WithMqttKeepAlive(0) footgun: paho stores keepalive as whole
  seconds, so sub-second truncates to 0 and 0 disables keepalive entirely.
- Add an internal test that calls initCl... (continued)

Coverage Stats

34 of 34 new or added lines in 2 files covered. (100.0%)

5 existing lines in 1 file now uncovered.

822 of 964 relevant lines covered (85.27%)

38.17 hits per line