lakestatus

What it is for

lakestatus is a command on the server that prints one screen: are all services healthy, how loaded is the machine, and what is in the lake. It only reads; it changes nothing. It is written in Python with the standard library only, so it needs no installation beyond one file.

What the screen shows

Section	Content
SERVICES	The systemd state of each service plus a real probe of its port, with a green, yellow or red dot. PostgREST is marked “old” and not counted
MACHINE	CPU load, memory and disks as bars
DATA	Counts of vendors, projects, folders and files; files by status with a bar; a 14-day upload sparkline; the last upload; views and live APIs; data-API and vendor-push calls in the last 24 hours

Use it

lakestatus                # one picture, then exit
lakestatus --watch        # redraw every 5 seconds until Ctrl-C
lakestatus --plain        # ASCII only, no colour (logs, e-mail, old terminals)
lakestatus --json         # the raw numbers instead of the picture
lakestatus --demo         # made-up numbers: try the layout anywhere, no server needed

The exit code is 0 when every service is healthy and 1 when something needs attention, so it works in cron or a monitoring check:

lakestatus --plain || echo "lake needs attention" | mail -s "lake" ops@example.com

Where it lives

Command	/usr/local/bin/lakestatus (step 17 copies lakestatus.py there)
Source	infra/lakestatus/lakestatus.py, tests in test_lakestatus.py
Reads	/opt/lake/cluster.env (role hosts and the registry password), the registry through psql, /proc, systemctl

The user running it must be able to read cluster.env (the admin user can; see step 19) and psql must be installed.

Install and test

bash infra/deploy.sh                                     # your machine
bash /opt/lake/scripts/install-all.sh 17                 # the server
python -m unittest discover -s infra/lakestatus          # tests (22), no server needed

The tests cover the drawing helpers (including the display width of emoji), settings parsing, probes against a local test server, the health rules, and rendering with an unreadable registry or an empty lake.

Not yet verified on the server itself

The systemd checks, /proc memory and uptime, psql as a subprocess, the Keycloak health port 9180 and the emoji rendering in every terminal were checked through the tunnel or in tests but not run on the server. Use --plain if your terminal draws the dots badly.