At a glance: open source, outbound-only monitoring

What it does

banshee checks at regular intervals that your equipment and services respond the way they should. It shows their state, and it sends an alert when one of them has a problem. That is what monitoring means.

It runs on your own servers, and the monitoring data stays there.

Who it is for

IT teams with an infrastructure to watch, from a handful of servers to tens of thousands of hosts, on one site or many.

For management, the question is usually how much of the team’s time monitoring eats up. Everything is set on screen, with no files or syntax to learn, so several people on the team can run it and monitoring no longer hangs on a single expert. The software itself costs no license fee, however many hosts you have.

How it works

banshee is made of two programs.

The master is the central server. It holds the inventory of what is monitored, the settings, the accounts and permissions, and it serves the web interface.
The pollers run the checks. You install them close to the equipment, in a data center, at a customer’s site or on a monitored server. They query the equipment over SNMP, SSH or HTTPS, judge the answers and send the results to the master.

The poller always opens the connection to the master, never the other way around. A remote site can therefore be monitored with no inbound port and no VPN.

Each check is called a probe. The screen shows the age of its last result, and a result that was not renewed in time is marked stale instead of staying green. Each poller also sends a heartbeat every minute; if it goes silent, its probes stop showing green.

An alert only goes out once the problem is confirmed, after the delay or number of failures you chose. It says why in one sentence, for example “eth0 at 93%, above 90% for 5 min”.

What it does not do

It does not change your equipment on its own. An SSH command or an HTTP request that would write to a device only goes out if a collector’s author wrote it into the code, and that code is reviewed before it is deployed.
The master does not take control of the pollers. It pushes no updates to them: a poller’s collectors are updated on the poller itself, by an administrator who sees the changes before confirming.
It has no dedicated high-availability mechanism for the master.
It does not handle dependencies between hosts. Each host is judged on its own.
No artificial intelligence decides the state of a probe. The rules are the ones the administrator ticked.
It has no paid edition, and no per-host or per-probe license.

License and price

banshee is free software, released under the GNU AGPL-3.0 license. It costs nothing, and there is only one edition, the same for everyone. The license gives you the right to read its code and adapt it; anyone who modifies it and makes it available to other users over a network must give them the code.

The project is supported by sponsors. It is also looking for partners able to deploy banshee and support the companies that adopt it.

Technical sheet

Central server: Linux amd64, Debian or Alpine, with a PostgreSQL database.
Pollers: Linux, FreeBSD, macOS (Intel and ARM), Windows Server 2016 and later. Windows Server 2008 and 2012 are monitored remotely, from a neighboring poller.
Collection protocols: SNMP v2c and v3, SSH, HTTP and HTTPS.
Collectors: written in Keen, banshee’s simple language, built into the software with no dependency on the system. They are pulled from repositories or written in the master’s editor.
Probe states: OK, WARNING, CRITICAL, MISCONFIGURED (the probe cannot measure because of its configuration), UNREACHABLE (the device returns no usable answer), PENDING (no result yet) and OLD (the last result is too old).
Target scale: 80,000 hosts and 600,000 probes. The inventory has been tested with 680,001 folders, hosts and probes.
License: AGPL-3.0.
Origin: designed and written in Alsace, France.

An administrator’s questions, from agents to high availability, are answered in the FAQ. The project’s story is on the About page, and where it stands is in the blog’s first post. To talk about your infrastructure, contact the team.