netreporter is (will be) a half-baked tool for monitoring basic connectivity of a network. It's aimed at recording data for understanding home network problems, but there's nothing specific to home networks.
Still vaporware.
First, this is a toy project, partly intended to learn Rust.
My goal is to see graphs of connectivity within my home network as well as from
my home network to WAN endpoints (e.g., 8.8.8.8) so that I can help understand
cases where the network is not working well. To make it concrete, I'd like to:
- easily stand up netreporter server instances on components to monitor
- support basic connectivity tests (e.g., ICMP ping, plus UDP and TCP analogs) among netreporter instances
- support basic short throughput tests over TCP among netreporter instances
Ideally, on my home network, I'd do this:
- set up a netreporter instance on computer A on the LAN (wired)
- set up a netreporter instance on an external endpoint B (across the WAN)
- run periodic connectivity checks between:
- A and the LAN router (ICMP ping)
- A and the LAN cable modem (ICMP ping)
- A and WAN endpoint (ICMP ping, netreporter-level UDP/TCP)
- A and B (ICMP ping, netreporter-level UDP/TCP)
- run periodic throughput checks between A and B (netreporter-level TCP)
The intent is to expose these all using Prometheus endpoints built into netreporter, then feed that into Grafana. With this set-up (which should be reasonably easy to orchestrate), one could identify windows of packet loss or high latency and determine whether it's related to the local network or the WAN connection.
I could also set up another endpoint on the LAN called C and run connectivity and throughput checks between A and C. In theory, one could set up multiple endpoints on different parts of the WAN to assess internet connectivity, but I don't have enough expertise to know whether this would be a reliable approach (and I'm not interested in that use-case).
It would be nice to be able to set up ad-hoc instances (e.g., from a laptop on wifi) so that when running into a new network issue you can start collecting data immediately in an organized way.
I want to support OS X and SmartOS, since those are what I run on endpoints.
netreporter is (will be) a server program that supports:
- configuration to periodically run tests against remote endpoints, including:
- ICMP ping checks against other ICMP endpoints
- a basic UDP ping check using other netreporter instances
- a basic TCP ping check using other netreporter instances
- a basic TCP throughput test using other netreporter instances
- listening for TCP connections and UDP packets from other netreporter instances to participate in any of the above TCP and UDP checks
- listening for TCP connections for reporting stats in Prometheus format
- build a command-line tool for sending an ICMP ping packet and getting a response
- build a pair of command-line tools for sending a basic TCP ping and responding to it (and reporting on the latency)
- add Prometheus support
- figure out how to expose metrics so that it's easy to get historical graphs
- prototype the whole system: Prometheus, Grafana, plus 1-2 netreporter instances
- add configuration for periodic runs of these various checks
- add basic TCP throughput test