Skip to content

Latest commit

 

History

History
104 lines (76 loc) · 3.5 KB

README.md

File metadata and controls

104 lines (76 loc) · 3.5 KB
Raw

catTracks

Tracking Cats and generating data

Troubleshooting

Check the status of the cattracks program instances and servers.

SSH to rotblauer.cattracks (find IP and SSH creds via Ansible). Run tmux a -t cats. This will show you a system log of the cattracks program.

Still no errors?

SSH to Freya. ps aux | grep trackermain. If Freya is down or this cattracks instance, the cat trackers won't be able to post their tracks. Freya forwards the tracks to the cattracks server, which again forwards tracks to the Google App Engine cattracks.cc server, which does Influx metrics and linestrings extraction and a few other experimental things.

20240124

  • rottor is done and gone. cattracks have move to smaller, dedicated servers -- ansible-cats.

  • cattracks anybatch2geojsonpoints - transforms trackpoint, geojson fc, [geojson features], ld geojson, csv... to []geojson.Point

  • api interfaces

  • simple gzip storage

  • track processors; cleaners, sifters, sorters. all of these should be AFTER original tracks storage. store them dirty.

  • tile makers

  • forward-populate is really nice. it could be improved and extended;

    • persistent caching
    • handle multiple targets

  • api metrics

    • tracks pushed
    • aqi queries, errors
    • tippe run

20231214

cattracks is too big, too heavy, and thus too expensive while master.json.gz is only 6GB master tippecanoe proc is the CPU+memory hog, b/c its tiling the whole world for all 250 million tracks every time tracks.db is the footprint hog, 355GB, storing all tracks uncompressed, and for no reason (we don't read or use the indexes, except for deduping points which almost-to-always is unnecessary)

things i want to change to make it lighter:

  • tracks.db
    • important things:
      • "catsnaps" bucket stores 3k+ tracks with snaps
        • these tracks have .notes.imgS3
        • the actual images are stores in S3/rotblauercatsnaps
        • /catsnaps handler serves the catsnaps
      • /lastKnown handler uses
        • bucket="stats",key=lastknown returns all cats last known location, indexed on catname
        • type LastKnown map[string]*trackPoint.TrackPoint
  • mbtiles generation
    • currently the procmaster tippe takes... a long time; 24hrs+, maybe even 48..72..96hrs+ (this is master.json.gz -> master.mbtiles)
    • i can run (w/ same tippe config) tippe on cat:uniqcells for all cats in 44m minutes on my laptop

so, what i want to do is:

  • get rid of places, they are disused and cluttery
  • don't store tracks in bolt except for ONLY track-snaps
  • save snaps locally, as well as S3
  • use a more meaningful key for the snaps, like catname_timestamp_lat_long.png

pre-202312

//TODO

  • LAT/LONG ✔️ db

  • ELEVATION ✔️ db

  • SPEED ✔️ db

  • compass heading ✔️ db

  • TILT

  • NUmber of cats

  • cat heartrate ✔️ db

  • plots, mapper ✔️

  • pics, first person and third person, upload if not black and with wifi and lots of battery

  • First generator on computer desktop, then IOS.

//TODO , need to grap GAE datastore, or maybe just forgetaboutit

plays well with https://github.com/rotblauer/bildungs-roamin

Starting development server

🍺

go run main.go

Frey bay bay

does this