What this guide covers
A WatchZit console checks devices it can reach. The devices of a branch office are often out of reach: they sit behind the office router's NAT, the VPN does not cover them, or the management network is closed on purpose. A remote collector solves this. It is the ordinary WatchZit agent on one computer at the site, marked as a collector in the console. It receives the checks of the devices of its site, runs them there, discovers the site's network and sends the results back over its own encrypted connection. When the line to the console is down, it keeps measuring, and a second collector can take over when the first one fails.
- What you need
- Prepare the server for outside agents
- Install the agent at the site
- Make it a remote collector
- Assign the devices of the site
- Discover the branch network
- A map per site
- Alerts when the branch line or the collector is down
- A second collector for failover
- What to keep in mind
- What to do next
What you need
- A running WatchZit console (the appliance or the install script; see the download page).
- At each site, one computer that stays on: Linux or Windows. A small VM or a Raspberry Pi-class machine is enough; the Linux agent exists for amd64 and arm64.
- From that computer, an outbound connection to the agent port 8444 of the server. Nothing has to be opened at the site: the collector connects out, and the console never connects in.
- Administrator rights in the console: only administrators make collectors and choose where devices are monitored from.
Prepare the server for outside agents
The console has two ports: 8443 for the web interface and API, and 8444 for agents. Keep 8443 for your admin networks and make only 8444 reachable from the branch offices, for example through a port forward on the firewall of the main office.
When the branches reach the server by another name than the one you use for the console, set it under Settings → Network and access as the Public agent address, for example https://monitor.example.com:8444. It goes into every enrollment code, so set it before you create codes. On the same page you can limit the networks agents may come from.
Install the agent at the site
- Under Agents, create an enrollment code: a name, how many days it is valid, how many computers may use it and the group new devices go to. The code is shown once. It contains the server address and the fingerprint of the agent certificate, so the agent trusts only your server.
- On a Windows computer, download the MSI from the Agents page and install it; it asks for the code. Silently:
msiexec /i watchzit-agent-x64.msi ENROLLCODE="WZ1-…" /qn. The service isWatchZitAgent. - On Linux, copy the command the Agents page shows: it downloads the agent from the agent port, checks its SHA-256 and installs it with the code as the systemd service
watchzit-agent. Or install the deb or rpm package of the release and runsudo watchzit-agent enroll --code "WZ1-…".
The computer appears as a device with its CPU, memory, disks and interfaces. If it does not connect, check that it reaches port 8444 and the allowed networks under Settings → Network and access; the agent log is in C:\ProgramData\WatchZit Agent on Windows and in journalctl -u watchzit-agent on Linux.
Make it a remote collector
Open the device page of the new agent and click Use as a remote collector in the Agent box. That is all on the collector side: it has no web interface of its own and nothing to configure locally — everything is set in the console.
The Agents page now lists it under Remote collectors, with its status and how many devices it monitors. A collector is also kept up to date like any other agent: an agent older than the server gets the server's version when it connects.
Assign the devices of the site
Every device has a Monitored from setting on its device page: This console or one of your remote collectors. For one device, choose the collector there and click Save.
For many devices at once, filter the device list (for example by the group of the branch), tick the devices and use Change selected: → Monitored from.
The collector then receives their checks — ping, TCP, SNMP, web pages and scenarios, certificates, DNS, databases, MikroTik RouterOS, JSON APIs — together with the passwords and SNMP settings they need, runs them on their intervals and sends the results to the console. The graphs, triggers and maps work exactly as for devices the console checks itself.
Discover the branch network
You do not have to add the branch's devices by hand. Under Devices → Discovery, add a range of the branch (for example 192.168.20.0/24; up to a /22) and choose the collector next to it. The collector scans the network of its own site: it pings every address, tries common ports, asks SNMP with your SNMP profiles and looks up DNS names. Run the scan now, every day or every week; with a schedule, new hosts that appear at the branch show up by themselves.
Tick the hosts you want and click Add as devices. Devices added from a range scanned by a collector are monitored from that collector automatically. The SNMP discovery on a device page, and the reading of facts, interfaces and VLANs, go through the collector too.
A map per site
A map for each branch keeps the picture clear. Under Maps, add a map, click Edit and put the branch's devices on it with Add device. Suggested links offers the cables WatchZit found on the devices, with the ports at both ends; Add all takes them, Arrange lays the map out by its cables, and Save keeps it. A floor plan can go behind it as a background picture.
Then make one overview map with a submap for each branch: a submap links to the branch map and shows its worst state, so a red branch stands out at a glance. For a wall screen, Settings → Kiosk displays shows maps full screen behind a secret link.
To watch the line itself, add a Network path (traceroute) check to a device the collector monitors: the collector traces from its own site, so you see each router on the way from the branch with its loss and latency, and when the route changed.
Alerts when the branch line or the collector is down
When the connection between the branch and the console breaks, the collector keeps running the checks. The results — up to a day, the newest 200,000 — wait on the collector and arrive when the connection is back, so the graphs have no gap.
Meanwhile the console does not pretend to know more than it does: the checks of the collector's devices show Unknown instead of down, because the site may be fine. The collector's own agent check opens a problem, and that is the message you get: one alert for the branch, not one per device. Where problems go is set under Settings → Notifications — email, webhook, Slack, Microsoft Teams, Telegram, ntfy, PagerDuty, SMS and more.
Inside the branch, dependencies keep the noise down. Open the branch map and click Dependencies: following the cables from the router, each device is offered the one before it. Click Set the ticked dependencies. When the branch router or a switch fails, the devices behind it are only unreachable and their problems are recorded but not sent. A No new data trigger catches checks that stop delivering values.
A second collector for failover
For an important site, install the agent on a second computer there and make it a collector too. Then choose it as Backup next to Monitored from on the device page, or for many devices in the device list under Change selected: → Backup collector.
While the first collector is offline, the backup runs the checks of its devices. When the first one is back, it takes them over again. To give the devices back to the console altogether, click Stop collecting on the collector's device page.
What to keep in mind
- Results kept during an outage do not survive a restart of the collector: after a restart it gets its checks from the console again, and what it had not yet sent is lost.
- The buffer holds up to a day, at most the newest 200,000 results. A longer outage leaves a gap.
- Hypervisor checks stay with the console; they are not run by a collector.
- Agentless Linux checks over SSH through a collector need the host key entered in the check: the collector cannot save it, so it shows the key it saw.
- While the line is down, the console shows the collector's devices as unknown; you see their real state only when the results arrive.
- Removing the agent on the device page makes its key unknown; to use that computer again, enroll it with a new code.
What to do next
- Read the agent and collector chapters of the documentation, and see everything WatchZit does on the features page.
- If the branches run MikroTik, follow A modern alternative to MikroTik The Dude for SNMP, the RouterOS API and drawing the maps.
- Get the appliance or the install script on the download page.
Questions and ideas are welcome in GitHub Discussions; companies can get commercial support from MaxZit.