Improving GeoIP

[jthiltges]

For geolocating caches and clients, we currently rely on IP addresses and use a combination of the MaxMind database and manually maintained list of overrides. Handling unlocated clients is an ongoing manual effort, and will likely continue as IPv6 usage increases. How can we improve the process? Some ideas from initial discussion:

• Storing GeoIPs in Registry, instead of every Director (for multi-directors) since it’s sort of stateful
• Using Cloudflare to augment the GeoIP lookup results https://derekweitzel.com/2022/01/22/improving-geoip/
• For cache/origins, using a location provided in config or as part of the registration
• In client jobs, send an identifier (such as GLIDEIN_Resource_Name) to Director to assist in geolocation

See also BrianB's comments on improving the algorithm for using GeoIP data: https://github.com/orgs/PelicanPlatform/discussions/1198