FAQ

I lost the password

Ask another administrator first: in Settings → Users, Edit on your account sets a new password and shows it once, and Reset 2FA on the same row clears a second factor that locks you out. That is the everyday answer — an instance with two admin accounts never needs the rest of this page.

When nobody can sign in any more, there is no recovery e-mail: start the server once with DUMBMONIT_RESET_PASSWORD=1, set in docker-compose.yml:

    environment:
      DUMBMONIT_RESET_PASSWORD: "1"

Then docker compose up -d.

Every account and every session is removed at startup and the UI shows /setup again. Create the first administrator, then start again without the line, and run docker compose up -d again. Monitors, rules, channels and the single sign-on settings are untouched.

No data after adding a monitor

Give it a minute: the first probe runs at the monitor’s interval (60 s by default) and writes reach VictoriaMetrics every 5 seconds (DUMBMONIT_FLUSH_INTERVAL_SECS). Then:

1. Open the monitor page and click Probe now. It reports how many samples and series one probe produced, or the exact error.
2. Misconfigured with an error under the address is a configuration problem: wrong community or password, refused certificate (enable “Accept an unverifiable certificate”), missing capability. It is shown, not alerted.
3. Unreachable with no error means the monitor did not answer at all. For SNMP, a wrong community looks exactly like a monitor that is off, because the monitor never answers it. Check the community, that SNMP is enabled, and that the monitor allows the server’s address.
4. Check that the container can reach the monitor: Docker networks and firewalls are the usual culprits.

Ping says “configuration error”

The ping monitor needs an ICMP echo socket. The container runs without any capability, so the Compose file allows those sockets with the net.ipv4.ping_group_range sysctl under the dumbmonit service: make sure the sysctls: lines are still there, then docker compose up -d. See ICMP ping without a capability.

The agent does not appear

• The installer does a test push before starting the service and stops with an explicit error if the URL or token is wrong. Fix /etc/dumbmonit/agent.yaml and restart dumbmonit-agent.
• The URL in the install command is the one your browser used to reach the UI. From the monitored machine, that URL must reach the server: curl https://server:8080/api/health from there tells you.
• Behind a reverse proxy, /api/ingest must be forwarded and bodies of up to 16 MB allowed (client_max_body_size 16m in nginx).
• journalctl -u dumbmonit-agent -f shows every rejected push. The token itself never appears in the log.
• A revoked token stops the agent at its next push; re-run the installer with a new token.

See Linux, macOS, FreeBSD and Windows agent.

Changing the port

The container listens on 8080; the host port is the left side of the mapping in docker-compose.yml. Edit it, then run docker compose up -d:

    ports:
      - "8099:8080"

To change the port inside the container, set DUMBMONIT_BIND, for example 0.0.0.0:9000, and adjust the port mapping.

Why is the alert “Advisory” when the rule says “warning”?

The API stores three severities: info, warning, critical. The interface displays them with the weather bulletin’s words: Info, Advisory, Warning. See Severities.

An alert is “Suppressed by parent”

The monitor has a parent monitor that is down (unreachable, or its availability check is failing), so its own alerts are held back rather than sent. Fix the parent; the child’s alerts resolve or fire on their own afterwards without a second notification. See Dependency suppression.

“Unusual CPU” never fires

It is a seasonal baseline that stays silent for its first 14 days on each series. During that time it only shows what it would have fired. See Baseline learning.

Can I use my own VictoriaMetrics?

Yes: set DUMBMONIT_VM_URL to its address (http://host:8428) and the embedded VictoriaMetrics is not started. Every metric is prefixed dumbmonit_, so it can share the instance with other tools. GET /api/health reports victoria.embedded: false once the switch is done.

Does DumbMonit need to be reachable from the internet?

No. It reaches out to monitors and to notification services; nothing calls back, except agents, which need to reach the server’s URL. Keep it on your LAN or behind a VPN; if you expose it, use a reverse proxy with a public certificate that talks HTTPS to DumbMonit (HTTPS and encryption).

Is there an API token?

Yes. Settings → API & assistants creates dmt_… tokens with a read or write scope; send one as Authorization: Bearer dmt_… on any route and no cookie or CSRF header is needed. The same token connects an assistant through MCP. Tokens never manage accounts, sign-in settings or other tokens. See the HTTP API reference.