How to run a health check
Choosing what to check, reading a finding properly, and getting a report you can hand to someone else.
The check reads the configuration and state of the devices you ticked, then judges each item in turn. It only reads, and it only advises: it tells you what it suggests, it never does it for you.
Run a check#
- Tick the devices you want checked.
- Choose Health check — from the main window, or straight from a device card in
- Choose which groups to include this time. The screen before it starts tells you how many checks you have
selected out of the total.
- Start the check.
It does not begin on its own. Deciding what to check this time is part of the job, not a step to skip.
Eight groups, 31 checks in total#
| This group | What it looks at |
|---|---|
| Version and firmware | The version currently running, compared against the release that fixed known problems; also whether the hardware's own firmware has kept up. It can optionally look the latest version up online |
| Ports and links | Ports that settled on a slower speed than they are capable of, ports that can only send one way at a time, error and dropped-packet counts, links that keep dropping in and out, and alarms reported by fibre modules themselves |
| PoE | On hardware that can power devices over the network cable, the power state of each port and anything needing attention |
| Time | Whether the clock is right and whether it is syncing |
| Resources and hardware | CPU, memory, storage, and the readings the hardware reports about itself |
| DHCP and WAN | The service that hands out addresses, and the state of the connection going out |
| Configuration validity | Settings that contradict each other, or that look set but will not actually do anything |
| Security | How exposed the device is from outside, and configuration problems a read-only session can spot |
Reading a finding#
Every row tells you three things: how serious it is, what it saw, and what it suggests you do.
| Level | Meaning |
|---|---|
| Normal | Checked, nothing to report. |
| Advice | Worth knowing; not a problem in itself. |
| Warning | Should be looked at. |
| Problem | Needs attention. |
Two more states matter just as much as the levels:
- Not applicable — this check means nothing on this device, for example a power check on hardware that
cannot supply power.
- Could not check — the data could not be read. This is not a pass. It is most often a limit of the
kind of connection you used; see below.
The summary counts those separately, so "no problems found" never quietly includes a pile of checks that were never read at all.
Always read the evidence. Every item writes out what it actually saw — the running version against the fixed version, the speed the port settled on against what it is capable of. A verdict with nothing behind it is not something you can act on.
Marking something you meant#
Some findings are deliberate — a link that really is at 100M because of what is on the other end, or a setting your environment requires.
Mark it as expected and write down why. The mark covers that one check on that one device, and it stays until you clear it. It is not a global mute: the same check on another device still reports normally.
Comparing with last time#
Results can be compared against a previous run, so a follow-up check shows what changed rather than making you re-read all of it. That is what makes a before-and-after pass around a maintenance window worth doing.
Exporting a report#
Export report writes the results as HTML or CSV.
Before it writes anything, it tells you what will be in the file: how many devices and rows, the scope, whether advice and ignored items are included, and whether device names are written as-is or replaced.
- HTML is for reading and handing over.
- CSV is for further processing. Advice gets its own column, and items marked as expected carry their own
flag and reason columns.
- The export covers the whole run. Search and filters you applied to the results view do not narrow it.
- If a scope would produce nothing — a "problems only" report on a clean run — it says so rather than writing
an empty file that reads like a clean bill of health.
What it will not do#
The check never changes a device. Every finding ends in a suggestion for you to decide on.
That is a deliberate line, not a missing feature: automatically "fixing" a network device by following a rule that cannot see why you set it up that way is how a maintenance window becomes an outage.
Next#
- Troubleshooting — when something comes back as "could not check".
- Running commands — acting on what the check found.