Provide metrics & status on the scionlab website

[cmeury]

Goals

• Provide insight into activity of the network for passers-by
• Operational status for more involved users

Ideas

• Bandwidth usage at individual APs
• Overall traffic on the network
• Number of ASes, infrastructure ASes, hosts, user ASes (active/total)
• Show overall health of the network, i.e. which infrastructure ASes are currently up/down etc. (maybe color-coding a version of the SCIONLab topology figure)
• Current commit SHA1 for SCIONLab, and date
• Status of the attachment points
• Status of the provided services (bwtestserver, echoserver)

Examples

• https://traffic.lan.switch.ch/
• Traffic Weather Map is a nice example, as it shows the current traffic status on the network: https://traffic.lan.switch.ch/pub/swiss-map/
• Interface graphs with historical data: https://traffic.lan.switch.ch/pub/graphs/geant2.html
• Aggregated traffic volume statistics to different networks: https://traffic.lan.switch.ch/pub/monthly/

[cmeury]

Quick research done:

• SWITCH uses: https://www.network-weathermap.com/ (also see community provided barebones Prometheus data source).
• Only suitable panel found for Grafana so far is novatec-sdg-panel.
• There's also an SVG plugin, which would involve quite a bit of effort, but maybe worth a look.

[AnotherKamila]

This is not entirely on-topic, because it is about alerts rather than metrics, but a long long time ago me and a certain awesome person hacked together evilham/prometheus-adlermanager. The interesting thing there is the whitelisting by labels, so that we can selectively only show some data to the public instead of exposing potentially sensitive information. Maybe that code could be useful? That way we might be able to create e.g. a "public" Grafana instance that through an Adlermanager-like proxy only has access to whitelisted metrics, or something like that.

EDIT: Um actually this was so long ago I can't find the relevant code XD But anyway, the point was that we might want to consider proxying Prometheus to only let the public see whitelisted metrics, and then use Prometheus-compatible tools to display/query/whatever.

(Also, but this is completely OT: do we want to deploy Adlermanager as a SCIONLab status page? The code was a result of a 48h hackathon, so it needs some love, but it could be in better shape with a day or two of work.)