#monitoring #uptime #alarming

bin+lib minmon

An opinionated minimal monitoring and alarming tool

16 unstable releases (5 breaking)

0.7.0 Feb 12, 2024
0.6.1 Dec 1, 2023
0.6.0 Oct 5, 2023
0.5.6 Jul 11, 2023
0.3.1 Dec 26, 2022

#87 in Unix APIs

Download history 23/week @ 2023-10-30 3/week @ 2023-11-06 1/week @ 2023-11-13 24/week @ 2023-11-20 100/week @ 2023-11-27 5/week @ 2023-12-04 18/week @ 2023-12-11 3/week @ 2023-12-18 35/week @ 2023-12-25 15/week @ 2024-01-01 2/week @ 2024-01-15 4/week @ 2024-01-22 34/week @ 2024-01-29 1/week @ 2024-02-05 175/week @ 2024-02-12

215 downloads per month


4.5K SLoC

MinMon - an opinionated minimal monitoring and alarming tool (for Linux)

This tool is just a single binary and a config file. No database, no GUI, no graphs. Just monitoring and alarms. I wrote this because the existing alternatives I could find were too heavy, mainly focused on nice GUIs with graphs (not on alarming), too complex to setup, or targeted at cloud/multi-instance setups.

test workflow docker workflow cargo-deny workflow dependency status
Latest SemVer release crates.io AUR version License


The checks read the measurement values that will be monitored by MinMon.


An action is triggered, when a check's alarm changes its state or a report event is triggered.


The absence of alarms can mean two things: everything is okay or the monitoring/alarming failed altogether. That's why MinMon can trigger regular report events to let you know that it's up and running.

Design decisions

  • No complex scripting language.
  • No fancy config directory structure - just a single TOML file.
  • No users, groups or roles.
  • No cryptic abbreviations. The few extra letters in the config file won't hurt anyone.
  • There are no predefined threshold names like "Warning" or "Critical". You might want more than just two, or only one. So that's up to you to define in the config.
  • The same check plugin can be used multiple times. You might want different levels to trigger different actions for different filesystems at different intervals.
  • Alarms are timed in "cycles" (i.e. multiples of the interval of the check) instead of seconds. It's not very user-friendly but helps to keep the internal processing and the code simple and efficient.
  • Alarms stand for themselves - they are not related. This means that depending on your configuration, two (or more) events may be triggered at the same time for the same check. There are cases where this could be undesirable.
  • Simple, clean, bloat-free code with good test coverage.
  • Depending on your configuration, there may be similar or identical blocks in the config file. This is a consequence of the flexibility and simpleness of the config file format.
  • All times and dates are UTC. No fiddling with local times and time zones.
  • No internal state is stored between restarts.
  • As of now it's only for Linux but it should be easy to adapt to other *NIXes or maybe even Windows.
  • Some of the things mentioned above may change in the future (see Roadmap).

Config file

The config file uses the TOML format and has the following sections:


System overview

graph TD
    A(Config file) --> B(Main loop)
    B -->|interval| C(Check 1)
    B -.-> D(Check 2..n)
    C -->|data| E(Alarm 1)
    C -.-> F(Alarm 2..m)
    E -->|cycles, repeat_cycles| G(Action)
    E -->|recover_cycles| H(Recover action)
    E -->|error_repeat_cycles| I(Error action)
    E --> J(Error recover action)

    style C fill:green;
    style D fill:green;
    style E fill:red;
    style F fill:red;
    style G fill:blue;
    style H fill:blue;
    style I fill:blue;
    style J fill:blue;

Alarm state machine

Each alarm has 3 possible states. "Good", "Bad" and "Error".
It takes cycles consecutive bad data points to trigger the transition from "Good" to "Bad" and recover_cycles good ones to go back. These transitions trigger the action and recover_action actions. During the "Bad" state, action will be triggered again every repeat_cycles cycles (if repeat_cycles is not 0).

The "Error" state is a bit special as it only "shadows" the other states. An error means that there is no data available at all, e.g. the filesystem usage for /home could not be determined. Since this should rarely ever happen, the transition to the error state always triggers the error_action on the first cycle. If there is valid data on the next cycle, the state machine continues as if the error state did not exist and the error_recover_action is triggered.

    direction LR

    [*] --> Good
    Good --> Good
    Good --> Bad: action/cycles
    Good --> Error: error_action

    Bad --> Good: recover_action/recover_cycles
    Bad --> Bad: repeat_action/repeat_cycles
    Bad --> Error: error_action

    Error --> Good: error_recover_action
    Error --> Bad: error_recover_action
    Error --> Error: error_repeat_action/error_repeat_cycles


Check the mountpoint at /home every minute. If the usage level exceeds 70% for 3 consecutive cycles (i.e. 3 minutes), the "Warning" alarm triggers the "Webhook 1" action. The action repeats every 100 cycles until the "Warning" alarm recovers. This happens after 5 consecutive cycles below 70% which also triggers the "Webhook 1" action. If there is an error while checking the filesystem usage, the "Log error" action is triggered. This is repeated every 200 cycles.


interval = 60
name = "Filesystem usage"
type = "FilesystemUsage"
mountpoints = ["/home"]

name = "Warning"
level = 70
cycles = 3
repeat_cycles = 100
action = "Webhook 1"
recover_cycles = 5
recover_action = "Webhook 1"
error_repeat_cycles = 200
error_action = "Log error"

name = "Webhook 1"
type = "Webhook"
url = "https://example.com/hook1"
body = """{"text": "{{check_name}}: Alarm '{{alarm_name}}' for mountpoint '{{check_id}}' changed state to *{{alarm_state}}* at {{level}}."}"""
headers = {"Content-Type" = "application/json"}

name = "Log error"
type = "Log"
level = "Error"
template = """{{check_name}} check didn't have valid data for alarm '{{alarm_name}}' and id '{{alarm_id}}': {{check_error}}."""

# This is a block comment. It demonstrates how to add another check and alarm.
# [[checks]]
# name = "System pressure"
# type = "PressureAverage"
# cpu = true
# avg60 = true
# [[checks.alarms]]
# name = "Warning"
# level = 80
# action = "Another action"

The webhook text will be rendered into something like "Warning: Filesystem usage on mountpoint '/home' reached 70%."


graph TD
    A(example.toml) --> B(Main loop)
    B -->|every 60 seconds| C(FilesystemUsage 1: '/srv')
    C -->|level '/srv': 60%| D(LevelAlarm 1: 70%)
    D -->|cycles: 3, repeat_cycles: 100| E(Action: Webhook 1)
    D -->|recover_cycles: 5| F(Recover action: Webhook 1)
    D -->|error_repeat_cycles: 200| G(Error action: Log error)

    style C fill:green;
    style D fill:red;
    style E fill:blue;
    style F fill:blue;
    style G fill:blue;

Some (more exotic) ideas

Just to give some ideas of what's possible:

  • Run it locally on your workstation and let it send you notifications to your desktop environment using the Process action and notify-send when the filesystem fills up.
  • Use the report in combination with the Webhook action and telepush and let it send you "I'm still alive, since {{minmon_uptime_iso}}!" once a week to your Telegram messenger for the peace of mind.


To improve the reusability of the actions, it's possible to define custom placeholders for the report, events, checks, alarms and actions. When an action is triggered, the placeholders (generic and custom) are merged into the final placeholder map. Inside the action (depending on the type of the action) the placeholders can be used in one or more config fields using the {{placeholder_name}} syntax. There are also some generic placeholders that are always available. Placeholders that don't have a value available when the action is triggered will be replaced by an empty string.


Filters can be applied to transform the measurement data. This has different use cases. For example:

  • Compensate for fluctuations in the measurement.
  • Determine the total network traffic over a number of cycles.

They can be configured for checks, in which case they affect all alarms that belong to the check, or alarms individually. Having both options reduces duplication in the config file in some cases. The check is the preferred place for filtering because it's only done once for all alarms which reduces memory and CPU usage.


Docker image

To pull the docker image use

docker pull ghcr.io/flo-at/minmon:latest

or the example docker-compose.yml file.
In both cases, read-only mount your config file to /etc/minmon.toml.

Build and install using cargo

Make sure cargo and OpenSSL are correctly installed on your local machine.
You can either install MinMon from crates.io using

cargo install --all-features minmon

Or if you already checked out the repository, you can build and install your local copy like this:

cargo install --all-features --path .

Copy the systemd.minmon.service file to /etc/systemd/system/minmon.service and place your config file at path /etc/minmon.toml. You can enable and start the service with systemctl daemon-reload && systemctl enable --now minmon.service.

Install from the AUR (Arch Linux)

Use your package manager of choice to install the minmon package from the AUR.
Place your config file at path /etc/minmon.toml. You can enable and start the service with systemctl daemon-reload && systemctl enable --now minmon.service.

systemd integration (optional)

Build with --features systemd to enable support for systemd.

  • Logging to journal.
  • Notify systemd about start-up completion (Type=notify).
  • Periodically reset systemd watchdog (WatchdogSec=x).

lm_sensors integration (optional)

Build with --features sensors to enable support for lm_sensors.
For the docker image, optionally mount your lm_sensors config file(s) to /etc/sensors.d/.
Note: libsensors is not cooperative and might theoretically block the event loop.




~358K SLoC