Network UPS Tools is a collection of programs which provide a common interface for monitoring and administering UPS, PDU and SCD hardware. It uses a layered approach to connect all of the parts.
Drivers are provided for a wide assortment of equipment. They understand the specific language of each device and map it back to a compatibility layer. This means both an expensive high end UPS, a simple "power strip" PDU, or any other power device can be handled transparently with a uniform management interface.
This information is cached by the network server upsd, which then
answers queries from the clients. upsd contains a number of access
control features to limit the abilities of the clients. Only authorized
hosts may monitor or control your hardware if you wish. Since the
notion of monitoring over the network is built into the software, you
can hang many systems off one large UPS, and they will all shut down
together. You can also use NUT to power on, off or cycle your data center
nodes, individually or globally through PDU outlets.
Clients such as upsmon check on the status of the hardware and do things
when necessary. The most important task is shutting down the operating
system cleanly before the UPS runs out of power. Other programs are
also provided to log information regularly, monitor status through your
web browser, and more.
NUT comes pre-packaged for many operating systems and embedded in storage, automation or virtualization appliances, and is also often shipped as the software companion by several UPS vendors. Of course, it is quite normal and supported to build your own — whether for an operating system which lacks it yet, or for an older distribution which lacks the current NUT version; whether to take advantage of new features or to troubleshoot a new UPS deployment with a debugger in hand.
Given its core position at the heart of your systems' lifecycle, we make it a point to have current NUT building and running anywhere, especially where older releases did work before (including "abandonware" like the servers and OSes from the turn of millennium): if those boxes are still alive and in need of power protection, they should be able to get it.
|
Tip
|
If you like how the NUT project helps protect your systems from power outages, please consider sponsoring or at least "starring" it on GitHub at https://github.com/networkupstools/nut/ - these stars are among metrics which the larger potential sponsors consider when choosing how to help FOSS projects. Keeping the lights shining in such a large non-regression build matrix is a big undertaking! See acknowledgements of organizations which help with NUT CI and other daily operations for an overview of the shared effort. |
As a FOSS project, for over a quarter of a century we welcome contributions of both core code (drivers and other features), build recipes and other integration elements to make it work on your favourite system, documentation revisions to make it more accessible to newcomers, as well as hardware vendor cooperation with first-hand driver and protocol submissions, and just about anything else you can think of.
If you are installing these programs for the first time, go read the installation instructions to find out how to do that. This document contains more information on what all of this stuff does.
When upgrading from an older version, always check the upgrading notes to see what may have changed. Compatibility issues and other changes will be listed there to ease the process.
Once NUT is installed, refer to the configuration notes for directions.
This is just an overview of the software. You should read the man pages, included example configuration files, and auxiliary documentation for the parts that you intend to use.
These programs are designed to share information over the network. In
the examples below, localhost is used as the hostname. This can also
be an IP address or a fully qualified domain name. You can specify a
port number if your upsd process runs on another port.
In the case of the program upsc, to view the variables on the UPS called
sparky on the upsd server running on the local machine, you’d do this:
/usr/local/ups/bin/upsc sparky@localhost
The default port number is 3493. You can change this with "configure --with-port" at compile-time. To make a client talk to upsd on a specific port, add it after the hostname with a colon, like this:
/usr/local/ups/bin/upsc sparky@localhost:1234
This is handy when you have a mixed environment and some of the systems are on different ports.
The general form for UPS identifiers is this:
<upsname>[@<hostname>[:<port>]]
Keep this in mind when viewing the examples below.
This package is broken down into several categories:
-
drivers - These programs talk directly to your UPS hardware.
-
server - upsd serves data from the drivers to the network.
-
clients - They talk to upsd and do things with the status data.
-
cgi-bin - Special class of clients that you can use with your web server.
-
scripts - Contains various scripts, like the Perl and Python binding, integration bits and applications.
These programs provide support for specific UPS models. They understand the protocols and port specifications which define status information and convert it to a form that upsd can understand.
To configure drivers, edit ups.conf. For this example, we’ll have a UPS
called "sparky" that uses the apcsmart driver and is connected to
/dev/ttyS1. That’s the second serial port on most Linux-based systems.
The entry in ups.conf looks like this:
[sparky] driver = apcsmart port = /dev/ttyS1
To start and stop drivers, use upsdrvctl of upsdrvsvcctl (installed on operating systems with a service management framework supported by NUT). By default, it will start or stop every UPS in the config file:
/usr/local/ups/sbin/upsdrvctl start /usr/local/ups/sbin/upsdrvctl stop
However, you can also just start or stop one by adding its name:
/usr/local/ups/sbin/upsdrvctl start sparky /usr/local/ups/sbin/upsdrvctl stop sparky
On operating systems with a supported service management framework, you might wrap your NUT drivers into individual services instances with:
/usr/local/ups/sbin/upsdrvsvcctl resync
and then manage those service instances with commands like:
/usr/local/ups/sbin/upsdrvsvcctl start sparky /usr/local/ups/sbin/upsdrvsvcctl stop sparky
To find the driver name for your device, refer to the section below called "HARDWARE SUPPORT TABLE".
Some drivers may require additional settings to properly communicate with your hardware. If it doesn’t detect your UPS by default, check the driver’s man page or help (-h) to see which options are available.
For example, the usbhid-ups driver allows you to use USB serial numbers to distinguish between units via the "serial" configuration option. To use this feature, just add another line to your ups.conf section for that UPS:
[sparky] driver = usbhid-ups port = auto serial = 1234567890
The Hardware Compatibility List is available in the source directory (nut-X.Y.Z/data/driver.list), and is generally distributed with packages. For example, it is available on Debian systems as:
/usr/share/nut/driver.list
This table is also available online.
If your driver has vanished, see the FAQ and Upgrading notes.
NUT provides several generic drivers that support a variety of very similar models.
-
The
genericupsdriver supports many serial models that use the same basic principle to communicate with the computer. This is known as "contact closure", and basically involves raising or lowering signals to indicate power status.This type of UPS tends to be cheaper, and only provides the very simplest data about power and battery status. Advanced features like battery charge readings and such require a "smart" UPS and a driver which supports it.
See the genericups(8) man page for more information.
-
The
usbhid-upsdriver attempts to communicate with USB HID Power Device Class (PDC) UPSes. These units generally implement the same basic protocol, with minor variations in the exact set of supported attributes. This driver also applies several correction factors when the UPS firmware reports values with incorrect scale factors.See the usbhid-ups(8) man page for more information.
-
The
nutdrv_qxdriver supports the Megatec / Q1 protocol that is used in many brands (Blazer, Energy Sistem, Fenton Technologies, Mustek, Voltronic Power and many others).See the nutdrv_qx(8) man page for more information.
-
The
snmp-upsdriver handles various SNMP enabled devices, from many different manufacturers. In SNMP terms,snmp-upsis a manager, that monitors SNMP agents.See the snmp-ups(8) man page for more information.
-
The
powerman-pduis a bridge to the PowerMan daemon, thus handling all PowerMan supported devices. The PowerMan project supports several serial and networked PDU, along with Blade and IPMI enabled servers.See the powerman-pdu(8) man page for more information.
-
The
apcupsd-upsdriver is a bridge to the Apcupsd daemon, thus handling all Apcupsd supported devices. The Apcupsd project supports many serial, USB and networked APC UPS.See the apcupsd-ups(8) man page for more information.
upsdrvctl can also shut down (power down) all of your UPS hardware.
|
Warning
|
If you play around with this command, expect your filesystems to die. Don’t power off your computers unless they’re ready for it: /usr/local/ups/sbin/upsdrvctl shutdown /usr/local/ups/sbin/upsdrvctl shutdown sparky |
You should read the Configuring automatic UPS shutdowns chapter to learn more about when to use this feature. If called at the wrong time, you may cause data loss by turning off a system with a filesystem mounted read-write.
NUT also provides an advanced support for power distribution units.
You should read the NUT outlets management and PDU notes chapter to learn more about when to use this feature.
upsd is responsible for passing data from the drivers to the client
programs via the network. It should be run immediately after upsdrvctl
in your system’s startup scripts.
upsd should be kept running whenever possible, as it is the only source
of status information for the monitoring clients like upsmon.
upsmon provides the essential feature that you expect to find in UPS
monitoring software: safe shutdowns when the power fails.
In the layered scheme of NUT software, it is a client. It has this separate section in the documentation since it is so important.
You configure it by telling it about UPSes that you want to monitor in upsmon.conf. Each UPS can be defined as one of two possible types: a "primary" or "secondary".
The monitored UPS possibly supplies power to this system running upsmon,
but more importantly — this system can manage the UPS (typically, this
instance of upsmon runs on the same system as the upsd and driver(s)):
it is capable and responsible for shutting it down when the battery is
depleted (or in another approach, lingering to deplete it or to tell the
UPS to reboot its load after too much time has elapsed and this system
is still alive — meaning wall power returned at a "wrong" moment).
The shutdown of this (primary) system itself, as well as eventually an UPS shutdown, occurs after any secondary systems ordered to shut down first have disconnected, or a critical urgency threshold was passed.
If your UPS is plugged directly into a system’s serial or USB port, the
upsmon process on that system should define its relation to that UPS
as a primary. It may be more complicated for higher-end UPSes with a
shared network management capability (typically via SNMP) or several
serial/USB ports that can be used simultaneously, and depends on what
vendors and drivers implement. Setups with several competing primaries
(for redundancy) are technically possible, if each one runs its own
full stack of NUT, but results can be random (currently NUT does not
provide a way to coordinate several entities managing the same device).
For a typical home user, there’s one computer connected to one UPS.
That means you would run on the same computer the whole NUT stack — a suitable driver, upsd, and upsmon in primary mode.
The monitored UPS may supply power to the system running upsmon (or
alternatively, it may be a monitoring station with zero PSUs fed by
that UPS), but more importantly, this system can’t manage the UPS — e.g. shut it down directly (through a locally running NUT driver).
Use this mode when you run multiple computers on the same UPS. Obviously, only one can be connected to the serial or USB port on a typical UPS, and that system is the primary. Everything else is a secondary.
For a typical home user, there’s one computer connected to one UPS.
That means you run a driver, upsd, and upsmon in primary mode.
More information on configuring upsmon can be found in these places:
-
The upsmon(8) man page
-
The stock
upsmon.confthat comes with the package
Clients talk to upsd over the network and do useful things with the data from the drivers. There are tools for command line access, and a few special clients which can be run through your web server as CGI programs.
For more details on specific programs, refer to their man pages.
upsc is a simple client that will display the values of variables known
to upsd and your UPS drivers. It will list every variable by default,
or just one if you specify an additional argument. This can be useful
in shell scripts for monitoring something without writing your own
network code.
upsc is a quick way to find out if your driver(s) and upsd are working
together properly. Just run upsc <ups> to see what’s going on, i.e.:
morbo:~$ upsc sparky@localhost ambient.humidity: 035.6 ambient.humidity.alarm.maximum: NO,NO ambient.humidity.alarm.minimum: NO,NO ambient.temperature: 25.14 ...
If you are interested in writing a simple client that monitors upsd,
the source code for upsc is a good way to learn about using the
upsclient functions.
See the upsc(8) man page and NUT command and variable naming scheme for more information.
upslog will write status information from upsd to a file at set
intervals. You can use this to generate graphs or reports with other
programs such as gnuplot.
upsrw allows you to display and change the read/write variables in your
UPS hardware. Not all devices or drivers implement this, so this may
not have any effect on your system.
A driver that supports read/write variables will give results like this:
$ upsrw sparky@localhost ( many skipped ) [ups.test.interval] Interval between self tests Type: ENUM Option: "1209600" Option: "604800" SELECTED Option: "0" ( more skipped )
On the other hand, one that doesn’t support them won’t print anything:
$ upsrw fenton@gearbox ( nothing )
upsrw requires administrator powers to change settings in the hardware.
Refer to upsd.users(5) for information on defining
users in upsd.
Some UPS hardware and drivers support the notion of an instant command - a feature such as starting a battery test, or powering off the load. You can use upscmd to list or invoke instant commands if your hardware/drivers support them.
Use the -l command to list them, like this:
$ upscmd -l sparky@localhost Instant commands supported on UPS [sparky@localhost]: load.on - Turn on the load immediately test.panel.start - Start testing the UPS panel calibrate.start - Start run time calibration calibrate.stop - Stop run time calibration ...
upscmd requires administrator powers to start instant commands.
To define users and passwords in upsd, see
upsd.users(5).
The CGI programs are clients that run through your web server. They allow you to see UPS status and perform certain administrative commands from any web browser. Javascript and cookies are not required.
These programs are not installed or compiled by default. To compile
and install them, first run configure --with-cgi, then do make and
make install. If you receive errors about "gd" during configure, go
get it and install it before continuing.
You can get the source here:
In the event that you need libpng or zlib in order to compile gd, they can be found at these URLs:
The CGI programs use hosts.conf to see if they are allowed to talk to a host. This keeps malicious visitors from creating queries from your web server to random hosts on the Internet.
If you get error messages that say "Access to that host is not authorized", you’re probably missing an entry in your hosts.conf.
upsstats generates web pages from HTML templates, and plugs in status
information in the right places. It looks like a distant relative of
APC’s old Powerchute interface. You can use it to monitor several
systems or just focus on one.
It also can generate IMG references to upsimage.
This is usually called by upsstats via IMG SRC tags to draw either the utility or outgoing voltage, battery charge percent, or load percent.
upsset provides several useful administration functions through a web
interface. You can use upsset to kick off instant commands on your UPS
hardware like running a battery test. You can also use it to change
variables in your UPS that accept user-specified values.
Essentially, upsset provides the functions of upsrw and upscmd, but
with a happy pointy-clicky interface.
upsset will not run until you convince it that you have secured your
system. You must secure your CGI path so that random interlopers
can’t run this program remotely. See the upsset.conf file. Once you
have secured the directory, you can enable this program in that
configuration file. It is not active by default.
The version numbers historically worked like this: if the middle number is odd, it’s a development tree, otherwise it is the stable tree.
The past stable trees were 1.0, 1.2, 1.4, 2.0, 2.2 and 2.4, with the latest such stable tree designated 2.6. The development trees were 1.1, 1.3, 1.5, 2.1 and 2.3. Since the 2.4 release, there is no real separate development branch anymore since the code is available through a revision control system (namely, Git — or actually Subversion back then), development happens in feature branches that are eventually merged into the main trunk, and its snapshots become published releases. As a result, subsequent versions (2.7 and 2.8) were released without regard for even/odd values of the minor version component.
Since 2.7 line of releases, sources are tracked in Git revision control system, with the project ecosystem being hosted on GitHub, and any code improvements or other contributions merged through common pull request approach and custom NUT CI testing on multiple platforms.
Major release jumps are mostly due to large changes to the features list. There have also been a number of architectural changes which may not be noticeable to most users, but which can impact developers.
Since NUT v2.8.2 or so, development iterations have additional version
components, to account for the amount of commits on the main branch
(master) since the last known Git tag, and amount of commits on the
developed feature branch that are unique to it compared to main branch.
This allows for a reasonably growing version of stable baseline and
local development, so that experimental packages can be installed as
upgrades (or well-exposed downgrades).
While the NUT releases retain the semantic versioning three-component standard, interim builds (trunk snapshots and development branches) can expose a much more complex structure with the amount of commits in the trunk since last release, and amount of commits on the branch unique to it (not in the trunk). Additional data may include overall amount of commits in the current build since last release, and the git commit has identifier of the built code base.
More details can be seen in docs/nut-versioning.adoc file in the
NUT source code base.
The network protocol for the current version of NUT should be backwards-compatible all the way back to version 1.4. A newer client should fail gracefully when querying an older server.
If you need more details about cross-compatibility of older NUT releases (1.x vs. 2.x), please see the Project history chapter.
If you are in need of help, refer to the Support instructions in the user manual.
Additional documentation can be found in:
-
the Developer Guide,
-
the Packager Guide.
The many people who have participated in creating and improving NUT are listed in the user manual acknowledgements appendix.
We would like to highlight some organizations which provide continuous support to the NUT project (and many other FOSS projects) on technological and organizational sides, such as helping keep the donations transparent, NUT CI farm afloat, and public resources visible. Thanks for keeping the clocks ticking, day and night:
The "NetworkUPSTools" organization on GitHub arranges a lot of things, including source code hosting for NUT itself and several related projects, team management, projects, issue and pull request discussions, sponsorship, nut-website rendering and hosting, some automated actions, and more… |
The Jenkins CI project and its huge plugin ecosystem provides the technological foundation for the largest island of the self-hosted NUT CI farm. There is a fair amount of cross-pollination between the upstream project and community, and the development done originally for the NUT CI farm. See more at Jenkins is the way to build multi-platform NUT article. |
Fosshost provided virtual machines where the multi-platform NUT CI farm with a jenkins-dynamatrix setup runs to arrange builds in numerous operating environments and a lot of toolkit versions and implementations. Some workers running on NUT community members' machines can also dial in to provide an example of their favourite platforms. Literally hundreds of NUT builds run for each iteration, to make sure NUT can always build and work everywhere. This allows us to ensure that NUT remains portable across two decades' worth of operating systems, compilers, script interpreters, tools and third-party dependencies. |
The CircleCI NUT pipeline allows us to test NUT CI builds on MacOS. |
The AppVeyor NUT pipeline allows us to test NUT CI builds on Windows (and publish preview tarballs with binaries). |
The DigitalOcean droplets allow us to host NUT CI farm Jenkins controller and the build agents for multiple operating systems. |
Gandi.Net took up the costs of NUT DNS hosting. |
https://opencollective.com/networkupstools allows us to arrange monetary donations and spending, with public transparency of everything that happens. |