WatchZit

An open-source Zabbix alternative: from install to useful monitoring

If you are looking for an alternative to Zabbix, this guide shows how WatchZit gets from an empty server to monitoring that alerts on the right things — with discovery, ready-made triggers, triggers that learn what is normal and maps that draw themselves, before you write a single template. It also says plainly what WatchZit does not do, and how to try it next to the system you run today.

What this guide covers

WatchZit is a free, open-source (AGPLv3) monitoring system for networks and servers: checks, graphs with years of history, triggers and notifications, dashboards and reports, and a live network map, in one web console on your own server. This guide does not compare the two products feature by feature. It shows what WatchZit does out of the box, where you still have to set things up, and where its limits are.

What you need

  • A virtual machine on VMware, Proxmox VE or Hyper-V with 2 CPUs, 2 GB of memory and a 16 GB disk — or a 64-bit Debian 12+, Ubuntu 22.04+, AlmaLinux or Rocky Linux 9+ server with PostgreSQL 13+.
  • SNMP read access (a v2c community or a v3 user) to the routers, switches, printers and UPSs you want to see in detail.
  • For servers, either the right to install the small WatchZit agent, or a user for SSH (Linux) or WinRM (Windows).

Install and sign in

Import the ready-made appliance from the download page — OVA for VMware, qcow2 for Proxmox VE and KVM, VHDX for Hyper-V (generation 1) or an ISO — or run the install script on a Linux server of your own:

curl -fsSL https://github.com/Max-Zit/watchzit/releases/latest/download/install.sh | sudo sh

The script installs PostgreSQL, checks the signature of the release, creates the database and prints the console address and the setup token. Open https://<server>:8443, enter the setup token, create the first administrator and set up two-factor sign-in with an authenticator app. The console answers only to private networks by default.

Find the devices

Create an SNMP profile first, under Settings → SNMP (v2c with a community, or v3 with SHA-2 and AES). Then go to Devices → Discovery and add a range such as 10.0.0.0/24 (up to a /22). WatchZit pings every address, tries common ports, asks SNMP with your profiles for the name, description and maker, and looks up the DNS name. Tick the hosts you want and click Add as devices: each gets a ping check (or a TCP check when ping is blocked) and the SNMP profile that answered.

Scan now, every day or every week. With a schedule, new hosts that appear on the network show up in the list by themselves.

What SNMP reads by itself

When a device has an SNMP profile, SNMP discovery offers what it found; tick what to watch. Loopbacks and interfaces without link are left out. What comes back depends on the kind of device:

  • Switches and routers: interfaces with link state and traffic, CPU, memory and storage; their VLANs with tagged and untagged ports, read once an hour; the VRRP role (master or backup).
  • Printers: toner, ink, drums and waste boxes as a level in percent.
  • UPSs (UPS-MIB or the APC PowerNet MIB): battery charge, runtime left and load.
  • MikroTik: temperatures, voltages, fans and power, each with its unit.
  • Anything else: any numeric OID as a check of its own.

Every value is kept and graphed: raw for 30 days, as 5-minute values for 90 days and as hourly values for two years by default (Settings → Metric data).

Triggers that are already there

A new console already has triggers under Problems → Triggers that watch every device:

TriggerWhen
Device downall checks of a device are down for 1 minute (critical)
Check downone check is down for 1 minute
High CPU load, Memory almost fullabove 90 % for 5 minutes
Disk almost full, Disk fullabove 90 % and 97 %
Disk full within two weeksthe trend of an agent disk's last week reaches 98 % within 14 days
Packet lossabove 20 % for 5 minutes
Certificate expires soon, Certificate expiresfewer than 21 days, and fewer than 7 (critical)

For what SNMP discovery finds there are ready-made triggers too: Toner or ink low below 10 % on printers and UPS runtime short below 10 minutes on UPSs. You only have to say where problems go: under Settings → Notifications set up email, Slack, Microsoft Teams, Discord, Telegram, ntfy, PagerDuty, SMS or a signed webhook.

To keep the noise down, set Depends on on devices — or open a map, click Dependencies and then Set the ticked dependencies, which follows the cables from the router. When the router of a site is down, you get one problem for the router; the devices behind it are only unreachable. Maintenance windows and acknowledgements do the rest.

Thresholds you do not have to guess

A fixed threshold is easy for a disk and hard for traffic, response times or the number of connections, which are high on Monday morning and low on Sunday night. WatchZit has two kinds of triggers for that:

  • Value unusual for the time (learned) needs no threshold. It learns from the last four weeks what is usual for the value at each hour of the week and opens a problem when the value is far from that — higher, lower or either. Sensitivity sets how far. It starts to compare after one week of history, and the problem names the usual value and the normal range for that hour.
  • Value will reach a threshold (forecast) draws a straight line through the last week of a value and opens a problem when that line reaches the threshold within the days you set — a disk that will be full in ten days, a temperature that keeps rising. The problem gives the value now, how fast it changes a day and the date it gets there.

A No new data trigger catches checks that stop delivering values, such as an agent that stopped reporting or a device that no longer answers SNMP.

Servers: agent or no agent

The WatchZit agent for Windows (MSI) and Linux (deb, rpm; x86-64 and arm64) reports CPU, memory, every disk, interfaces, services you choose, watched processes, Hyper-V machines and Docker containers, and sends errors from the Windows event logs and the systemd journal to Logs. It makes one outbound connection to port 8444, is enrolled with a one-time code (WZ1-…) from Agents and updates itself from the console. Its values get graphs and the built-in triggers by themselves.

Where you may not install anything, add the check Linux over SSH (no agent) or Windows over WinRM (no agent). They read CPU, memory, swap or the page file, the fullest disk and the uptime, with the same value names as the agent, so the same triggers and forecasts work. The guide Agentless Linux and Windows server monitoring shows the setup step by step.

Maps that draw themselves

Every 15 minutes WatchZit reads from the devices where things are plugged in: LLDP, CDP and MikroTik neighbour tables, the MAC address tables of switches, port descriptions and saved configurations. Under Maps, add a map, click Edit, put a few devices on it and open Suggested links: it lists the cables with the ports at both ends. Add all takes them. Turn on Draw by itself and Also add the devices they lead to, and the map keeps adding new links and the devices behind them. Arrange lays the map out by its cables; Save keeps it.

On the map, devices are coloured by their state and cables by their traffic, with pictures of some 3,500 real device models. For a wall screen, Settings → Kiosk displays shows maps full screen behind a secret link. The guide A modern alternative to MikroTik The Dude goes through maps in detail.

Templates, when you want them

WatchZit has no built-in templates, and for the basics it does not need them: the SNMP discovery and the agent find interfaces, disks, CPU and memory themselves, and the built-in triggers apply to all devices. Templates are for your own standards — for example the same web, certificate and TCP checks on every web server:

  1. Under Templates, create a template, choose the kinds of devices it is for and add its checks (ping, TCP, web page, TLS certificate, DNS and SNMP checks).
  2. Use Apply to for all devices of its kinds or a group, or Apply template on one device page. Applying only adds the checks a device does not have yet, so applying twice does no harm.
  3. Tick Apply to new devices of these kinds by itself, and devices added later by hand or from discovery get the checks at once.

For many devices at once, filter the device list, tick the devices and use Change selected: to set the group, the SNMP profile, the dependency or a template.

Dashboards, data sources, reports and the API

  • Dashboards show graphs, current values, device states, open problems and map tiles, and reload every minute for a wall screen.
  • Data sources (Settings → Data sources) bring time series from PostgreSQL, MySQL/MariaDB, Microsoft SQL Server, Prometheus, InfluxDB or a JSON service onto graphs and dashboards next to WatchZit's own values.
  • Reports show the availability of each device for 7, 30 or 90 days or a month, with SLA targets per group; they can go out by email every week or month with a CSV attached.
  • Syslog and SNMP traps (Settings → Syslog and traps) and Traffic flows (NetFlow v5/v9, IPFIX, sFlow) are received by the same server.
  • A REST API with tokens reads devices, checks, graphs and problems, acknowledges problems and plans maintenance.

What WatchZit does not do

  • No importer from other monitoring systems: devices come from discovery, the agent or by hand, and history starts on the day you add them.
  • No built-in templates. Templates hold ping, TCP, web, certificate, DNS and SNMP checks; checks that need a password are added per device.
  • No check that runs your own scripts. Values of your own come from an SNMP OID, a JSON API value check against a service of yours, or a data source on a graph.
  • Learned triggers need history: they compare only after one week, and forecasts need at least six hours of a value.
  • The agentless checks cover the basics (CPU, memory, swap or page file, the fullest disk, uptime). Every drive, interfaces, services, processes and logs need the agent.
  • The documented API reads data, acknowledges problems and plans maintenance; devices, checks and triggers are set up in the console.
  • The server runs on Linux (x86-64) with PostgreSQL, or as the appliance.

Run WatchZit next to your current monitoring

You do not have to switch in one day. WatchZit runs on its own VM and changes nothing on the devices it watches, so both systems can run side by side for a few weeks:

  1. Install WatchZit on a separate VM and give it its own read-only SNMP community or v3 user, so you can tell its queries apart and remove its access later.
  2. Start with one site or one group: discover it, set dependencies and send WatchZit's problems to a test channel first, not to the on-call team.
  3. Install the WatchZit agent next to whatever runs on the servers today, or use the agentless checks; the agent needs only its outbound connection to port 8444.
  4. If your devices can send syslog, SNMP traps or flows to a second destination, add WatchZit as one (UDP 514, 162, 2055 and 6343 by default) instead of moving them over.
  5. Compare for a few weeks which problems each system raised and which ones mattered, then decide.

What to do next

Questions and ideas are welcome in GitHub Discussions; companies can get commercial support from MaxZit.

Zabbix is a trademark of its owner. MaxZit is not affiliated with, sponsored or endorsed by the owner of Zabbix; the name is used only to describe what this guide is about.