FAQ
Why not just use one of the other monitoring tools?
If you already run something you are happy with, there is no reason to switch — this is not an argument against anything else out there. ProbeGlass exists for the case where you want the monitoring to be smaller than the thing it monitors: one binary, one file, about three megabytes of RAM, and a dependency tree you can read in an afternoon. The introduction has the comparison in a table.
Does it need a database server?
No. It uses SQLite, through a pure-Go translation of it, so there is nothing to install alongside it. The database is one file you can copy.
Can I run it on a Raspberry Pi?
Yes. The image is published for linux/arm64 and linux/arm/v7. The arm64 image
is cross-compiled rather than emulated, so it builds at native speed.
There is no separate binary for a Pi: the container image is the only thing ProbeGlass publishes.
Why does it not need NET_RAW for most things?
Only ICMP ping needs it, and only the raw-socket fallback at that. The ping sensor
tries an unprivileged datagram ICMP socket first, which works on Linux and macOS
without any capability as long as net.ipv4.ping_group_range permits the
container's group. If neither socket works, the sensor says exactly which one it
tried and what the kernel replied — the most confusing failure in containerised
monitoring is a ping that silently does nothing.
How much disk does it use?
Roughly 60 bytes per raw probe result. One sensor every 60 seconds is about 86 KB a day, 600 KB a week. After the raw retention window (7 days by default) only the hourly and daily rollups remain, and they are far smaller.
Can I keep history for longer than 90 days?
Daily rollups are kept forever, so a range of a year or more is already answerable
— it is read from the daily table. retention.hourly controls how long the
mid-resolution data survives; retention.raw controls the fine detail. See
History and uptime.
Why is a failed probe shown as "Flaky" and not "Down"?
Because at that moment the honest answer is "the last probe failed, and it has not
failed often enough yet to call it down". With retries: 3, one failed probe is a
blip. Calling it Down would be a lie, and calling it Up would be a bigger one — it
would hide a sensor that is starting to go. It becomes Down when the failure streak
reaches the threshold, and that is when the incident opens and the alert goes out.
Flaky is a yellow pill and nothing else: no incident, no alert. Unknown is the state a sensor wears before its very first probe, and it is the one that means "we have never heard from this at all".
An alert says "was down for 0s". Is that a bug?
It was, and it is fixed. The recovery alert now names the real outage length, computed from the incident's start time. If you see it again, please open an issue with the alert text.
Can I use it for a public status page?
Not yet. A read-only public view on its own path and its own token is on the roadmap. Today, the dashboard is authenticated, full stop.
Can I scrape it from another monitoring stack?
Not directly — there is no /metrics endpoint yet, and it is on the roadmap. In
the meantime the JSON API is scrapeable with a small exporter, and the generic
webhook can push every event into whatever you already run.
Multiple users?
Yes, up to a point. A fresh install creates one administrator; under Settings → Users you add as many more as you like, as operator or viewer, and any of them can enrol a second factor. What they all share is the same set of sensors, the same history and the same incidents — ProbeGlass is one homelab, and multi-tenancy is explicitly out of scope. Running two containers is the answer to wanting two.
What each role may do is in Security.
Can I import an existing monitoring configuration?
Not with a tool, no. There is no importer yet — it is on the roadmap. The practical route today is to declare your sensors in a YAML file, which is a better place for them anyway, and which the file-wins rule will then keep in sync on every start.
Where is the data, and how do I back it up?
/data/probeglass.db — one SQLite file. Stop the container and copy it. See
Backup and restore.
The dashboard shows a spinner and never loads
There is now a watchdog that replaces the spinner with a real message after twelve seconds, and a reload button. See Troubleshooting.
Something is wrong inside the container, and there is no shell
Run docker exec probeglass probeglass doctor. The image is distroless — no
shell, no curl, no tools — so that is the one command you have, and it reports the
configuration, the data directory, the database's size, integrity and schema, the
stored sensors and channels, and whether the daemon is answering. It opens the
database read-only and never sends an alert, so it is safe to run on a database you
are worried about. See Command line.
How do I reset a forgotten password?
Another administrator can do it under Settings → Access → Users. If that one account is
the one you have lost, there is no recovery endpoint by design — a "reset the
password" route is a way in for anyone who can reach the port. Delete the rows from
the users table of the SQLite file and the setup screen returns on the next
request. See Installation.
Can I contribute a translation?
Yes, and it is a small job. The dashboard has no user-facing strings in its markup — everything goes through a message catalogue — so a new language is one dictionary file and one entry in the picker list. German is the first planned addition. See Contributing.
Is there a hosted version?
No. ProbeGlass is a homelab tool: you run it, its data stays with you, and there is nothing to sign up for.
Who makes this? Is it a company?
No. ProbeGlass is a hobby project by somebody who is not a professional software developer — one person, spare time, written for their own homelab, and shared on the chance that somebody else finds it useful. There is no company, no support contract and no warranty — it is MIT licensed and comes "as is", and nobody is liable for anything it does, including a missed alert. If you are weighing that up, read Who makes this, and how before you decide.