When something is wrong
Start with sudo reportlog doctor. It checks settings, mail, the database
and its migrations, that PostgreSQL listens on no network port, the API, the
site and its certificate through Caddy, backups local and off-site, every
timer and the last result of every job, unattended updates, disk space and
the Node version. Each line says what to do.
| Symptom | Look at |
|---|---|
| The site is down | systemctl status reportlog-api caddy, journalctl -u reportlog-api -n 50 |
| No certificate | journalctl -u caddy: DNS not pointing here, or 80/443 closed |
| TLS error on www. | Add the name to SITE_ALIASES in deploy.env, then sudo reportlog deploy |
| Mail not arriving | sudo reportlog mail-test you@example.org |
| Owner locked out | sudo reportlog reset-2fa <the owner's address> (Two-factor login, above) |
| A job failed | journalctl -u reportlog-job@<name> -n 50 |
| An error, in code | Bugsink, if ERROR_TRACKING_DSN is set; otherwise the journal |
| Disk filling | du -sh /srv/reportlog/releases/* /var/backups/reportlog |
Logs never contain an IP address or a full URL: Caddy keeps no access log,
and the API logs route patterns. Caddy does pass the visitor's address to the
API (X-Forwarded-For, with TRUST_PROXY=1), which uses it only for the
sign-in and form limits it counts in memory, under keyed hashes; it is never
written down.