Troubleshooting

Start with baba health: it names the part that's broken. logLevel: "debug" (or BABA_LOG_LEVEL=debug) logs every reading, breach count and incident decision.

Config

no config file found at "…", run 'baba setup' to create one

There's no config.json at the path baba reads (--config, else $BABA_CONFIG_PATH, else /var/lib/baba/config.json), no config.default.json next to it, and the environment doesn't provide a notifier. Run baba setup, or pass --config with the right path.

A downloaded binary run as ./baba start still looks in /var/lib/baba, not in the current directory: use ./baba start --config config.json.

invalid config

The message lists every problem with its key:

baba: invalid config:
  • checks.cpu.usageThresholdPercent: Must be between 0 and 100

Fix them in config.json, or rerun baba setup. A JSON syntax or type error (a string where a number goes) is reported by the JSON parser instead. Unknown keys are ignored, so a misspelled key keeps its default rather than failing: compare with Configuration file.

invalid value for BABA_…

An environment variable meant as a number doesn't parse. The *_CONSECUTIVE_BREACHES ones must be whole numbers.

Alerts

Nothing arrives

1. baba validate: if it fails, the error names the notifier and the HTTP status or network error.
2. Discord HTTP 401 or 404: the webhook was deleted or the URL is truncated. Create a new one.
3. Telegram HTTP 400 with chat not found: wrong chatId, or the bot isn't in that chat. HTTP 401: wrong botToken.
4. Still nothing on a real breach: with logLevel: "debug", look for breach lines. The alert only fires after consecutiveBreaches breaches in a row, and a reading equal to the threshold isn't a breach.

See Incidents for the rules.

Telegram shows ** around values

Telegram messages are plain text; the stars are Discord's bold.

An alert failed and never came back

There's no retry. The incident records the notification with ✗ (baba get incident <id>), and the next reminder, after reminderIntervalMinutes, is the next attempt.

Checks

Disk: no volumes found

None of checks.disk.volumes matches a mount point or device exactly as the mount table lists it. Check with findmnt (Linux) or mount (macOS). In Docker, the container only sees its own mounts: see Disks in a container.

No temperature in the status line

No sensor was found. On Linux, check ls /sys/class/hwmon/*/name shows one of the CPU sensors listed in Checks, and that /sys is mounted in Docker. On macOS there's no CPU source; only NVIDIA GPUs report.

GPU: N/A

nvidia-smi isn't in the PATH, fails, or reports no memory figures. Run nvidia-smi yourself. Non-NVIDIA GPUs aren't supported.

✗ system: cannot read memory (…), is --pid=host set?

/proc/meminfo (Linux) or sysctl/vm_stat (macOS) couldn't be read. In Docker, run with --pid=host and --privileged.

Service

✗ incidents: not found at "…", has the service started at least once?

The history file is created by the first baba start. On a fresh install, start the service; if it's running, check database.path points where the service writes.

No sudo access, installing as a user service instead

sudo cp to /etc/systemd/system failed, so the unit went to ~/.config/systemd/user. User services stop when you log out; run loginctl enable-linger "$USER", or rerun baba install from an account with sudo.

no log file at /var/lib/baba/baba.log

Only the background service writes that file. In the foreground, logs go to the terminal; in Docker, use docker logs.

moving the old SQLite incident database aside

The TypeScript baba's baba.db was found at database.path. It's renamed to baba.db.sqlite.bak and the history starts fresh; see Incidents.

Updates

could not reach GitHub releases

No network, or GitHub's API rate limit (60 requests an hour per IP without a token). Try again later, or download the release by hand.

download failed: HTTP 404

The latest release has no binary for this OS and architecture. Only Linux and macOS on x64 and arm64 are built.