feat(coderd): add prometheus metrics to servertailnet #11988
Conversation
|
This stack of pull requests is managed by Graphite. Learn more about stacking. |
| } | ||
|
|
||
| func (c *instrumentedConn) Close() error { | ||
| c.closeOnce.Do(func() { |
There was a problem hiding this comment.
are network connections always explicitly closed?
There was a problem hiding this comment.
This one I'm not 100% sure on. Admittedly, I found this idea from a stackoverflow post which seemed to work for a couple other people. Was planning to get this into dev and monitor to make sure it works as intended with a lot more usage than I can reproduce myself.
There was a problem hiding this comment.
I guess maybe they should be and this gauge will tell us if there's a leak...
coderd/tailnet.go
Outdated
| connsPerAgent: prometheus.NewGauge(prometheus.GaugeOpts{ | ||
| Namespace: "coder", | ||
| Subsystem: "servertailnet", | ||
| Name: "open_tcp_connections", |
There was a problem hiding this comment.
I realize we only use servertailnet for HTTP proxying at present, but what do you think about changing this to open_connections and making the network a label (e.g. network=tcp? Like, I dunno, what if some customer wants us to proxy QUIC or some shit?
There was a problem hiding this comment.
Yeah that's a good point. If we were to keep it like this we would definitely preclude ourselves from expanding this without a breaking change. Adding a network=tcp with it really only ever being tcp for the forseeable future basically has the same implications and usability as not having it, probably a good change.
5f9a1d5 to
02a7089
Compare
No description provided.