Troubleshooting

speedtest binary not found

The Docker image bundles the Ookla CLI at /usr/local/bin/speedtest, so this only affects source builds. Install the Speedtest CLI and make sure speedtest is on the $PATH of the environment running the app. Binary resolution happens only inside the runner. A run that fails immediately with a not-found error is this. (The TCP-connect ping runner needs no binary.)

Database connection errors

Confirm DATABASE_URL is correct and reachable, and that migrations have run:

mix ecto.migrate

Under Docker, check that the Postgres container is up and on the same network as Baudflow. If ECTO_IPV6=true is set, make sure Postgres actually listens on IPv6.

A ping schedule reads 100% packet loss

The ping runner opens TCP connections to host:port and treats a completed handshake as a latency sample. If the target doesn’t accept TCP on the configured port, every connect fails and reads as 100% loss. Defaults target 1.1.1.1:443 (Cloudflare), which always speaks TCP. Point a schedule at a different target_host / target_port (or ping_target / ping_port globally) at something that accepts the connection.

Tests pass locally but fail in CI

CI runs the full gate: mix lint plus the suite against a fresh Postgres service container. just precommit is a fast loop that skips lint, so it can be green locally while CI is red. Run the CI-equivalent gate before pushing: just check.

Port 4000 already in use

Stop the conflicting process, or set a different PORT.

No data flowing

• Verify a schedule is enabled and its cron is valid; a bad cron on one row is logged and skipped, not fatal.
• Check the Oban queues. Workers stay decoupled, so a stuck queue in one stage won’t surface in another.
• Make sure the scheduler is enqueueing: look for SchedulerWorker jobs.

Disk filling up

Measurements store the full raw result JSON. Lower retention_days if history is growing faster than you want; the CleanupWorker prunes daily at 03:00.