Support / Install a probe
Install a probe
A SquidGrid probe is one lightweight container you run at a site. It measures
the paths you point it at and sends readings out — nothing listens, nothing captures
your traffic. Most people are done in a couple of minutes.
Before you start
- A host at the site that can run a container — Docker or Podman, Linux, x86-64
or ARM.
- Outbound connectivity only: HTTPS on TCP 443 to
collector.squidgrid.net
for enrollment and result delivery, outbound ICMP echo for reachability and latency,
and outbound UDP with returned ICMP for traceroute. Most networks allow all three by
default; restrictive egress policies or some container platforms may need explicit allowances.
- No inbound ports and no root. The probe runs unprivileged and listens for nothing.
- Very little of the host: a probe idles on a fraction of a core and a few tens of MB of RAM.
One probe per site. A probe measures from where it runs, so put one at each
location whose paths you want to see. Your plan includes a set number of sites; add more from
the
targets screen or your account panel.
Install it
- Open your Customer dashboard and go to Sites (or Add a site).
- Name the site and copy the enrollment command it shows you. The command already
contains this site's key — it is unique to that site and enrolls it to your account.
- Paste the command on the host at that site and run it. The probe pulls the
squidgrid/probe image and starts.
- Within a minute the site shows checked in on the dashboard. Measurements begin on
the next cycle.
The command your dashboard gives you is ready to paste. It looks like this — use the exact
line from the dashboard, which has your site's key filled in:
docker run -d --name squidgrid-probe \
--restart unless-stopped \
--cap-drop ALL \
-e SQUIDGRID_ENROLL=<your-site-key> \
squidgrid/probe
Podman is identical — substitute podman for docker. The image is
public on Docker Hub and you are welcome to inspect it before you run it.
Confirm it is measuring
On the dashboard the site moves to checked in, and its first readings appear once it has
run a few measurement cycles. If a site stays pending for more than a few minutes, it is
almost always outbound 443 being blocked — check that the host can reach the internet over
HTTPS. Nothing needs to reach in to the probe.
What the probe does and does not do
| Does | Does not |
| Measures the destinations you configure | Touch your internal network unless you point it there |
| Sends measurements only, outbound over HTTPS | Open any inbound port or listen for connections |
| Runs unprivileged, capabilities dropped | Run as root or need elevated privileges |
| Records latency, loss, jitter and the route | Capture, store or inspect your traffic |
The full detail of what the fleet sends and at what rate is on
how our probes measure, and the trust boundary is on the
security model page.
Updating and removing
To update, pull the image and re-run the same command — the site keeps its identity and
history. To remove a probe entirely, stop and delete the container
(docker rm -f squidgrid-probe). That is the whole uninstall: nothing is left on the
host, and the site simply stops reporting.
Next: Add targets & link sites
· Reading your dashboard
· back to all guides