TheWitness · Oct 16, 2026 · Oct 16, 2026 · Oct 16, 2026 · Oct 16, 2026 · Oct 16, 2026
diff --git a/CHANGELOG.md b/CHANGELOG.md
@@ -1,13 +1,22 @@
 # ChangeLog
 
---- develop ---
-
+--- 1.4.0 ---
+
+* feature: Add support for monitoring UPSes through Network UPS Tools (NUT) as a third collector type, "NUT Based". The poller talks to `upsd` (TCP 3493 by default) over a native PHP socket implementation of the NUT protocol, so no `upsc` or other CLI tool is needed. Optional username/password authentication and STARTTLS (with certificate verification) are supported, and one NUT server can serve several UPSes through the new NUT UPS Name setting. NUT status flags are translated to the existing apcupsd vocabulary (OL to ONLINE, OB to ONBATT, and so on) so the dashboards, fleet totals and alerts keep working unchanged
+* issue: Widen the status, model, serial number, firmware, driver and name columns of `apcupsd_ups_stats` so multi-flag NUT statuses and longer NUT strings are not truncated
+* feature: Let NUT STARTTLS work with self-signed certificates: a per-UPS CA / Certificate File to verify the server against (certificate and host name are still checked), or an Allow Self-Signed Certificate option that encrypts without verifying the server
+* feature: Add a per-UPS Timeout (seconds) setting, 5 by default and limited to 1-120, for `apcupsd` and NUT UPS'. NUT connects and conversations are now bounded by twice the timeout in total, and `apcaccess` is run through Cacti's `exec_with_timeout()` so a hung `apcaccess` can no longer stall the poller; a timed out `apcaccess` now records an error on the UPS
+* feature: Make the UPSes table on the Fleet Dashboard sort like the HMIB dashboards: click (or press Enter/Space on) a column header to sort and click again to reverse, in the card and in its maximize dialog
+* issue: Right-align the Load, Charge, Time Left, Est. Power and Last Updated columns of the Fleet Dashboard UPSes table
 * feature: Share the UPS Dashboard card arrangement across every UPS (graph cards stay per UPS since they belong to each device), hide N/A rows in the UPS Information card, combine Cable and Driver on one row, and add a per-UPS Estimated Power gauge card that reuses the shared nominal power x load % estimate
 * issue: Show N/A for a slave UPS's (e.g. ONLINE SLAVE) estimated power on the Fleet Dashboard and leave it out of the fleet power total, since its draw is already counted in its master's
-
 * feature: Estimate each UPS's power draw as nominal power x load % for UPSes that do not report real wattage, and show it on the Fleet Dashboard (an "Est. Power" tile, a new "Fleet Power" gauge card of total estimated watts against rated capacity, and an "Est. Power" column with a fleet total in the UPSes card) and as a sortable "Est. Power" column in the UPSes list
-* feature: Make the Fleet Dashboard UPSes table sortable by UPS, Status, Load, Charge, Time Left, Est. Power and Last Updated; the choice persists in the browser and survives card refreshes
+* feature: Make the Fleet Dashboard UPSes table sortable by UPS, Status, Load, Charge, Time Left, Est. Power and Last Updated
 * issue: Stop the dashboard auto-refresh interval following the user to another page by driving Cacti's `refreshMSeconds`/`refreshPage` page-timer globals (as the Graphs page does) instead of storing the interval in the session
+* feature: Auto-refresh now updates the dashboard's cards in place instead of reloading the whole page: through Cacti's `refreshFunction` page-timer hook where core has it (Cacti 1.3+), and with a timer of its own on Cacti 1.2.x, which has no such hook and stops when the dashboard leaves the screen. The chosen table sort is kept across a refresh (the live one, even if it was changed while the refresh was in flight), overlapping refresh rounds are skipped, and the full-page refresh remains the fallback when the dashboard script is not loaded
+* issue: Mark a NUT UPS's linked Cacti host down when its NUT query fails (these devices have availability checking disabled, so core polling cannot detect it), and stop `upsd` from holding a poll open past the deadline by trickling bytes
+* issue: Stop the UPS edit form from rewriting an existing UPS's port (3551 to 3493 or back) when it opens; the default port is now only swapped when the collector type is changed by hand
+* issue: Store the start time, manufacture date, master update, last on battery and last self test of `apcupsd_ups_stats` as unknown (NULL) when the UPS does not report them, instead of letting each poll invent the current time
 * security: Replace the confirmation pages' inline `onClick='cactiReturnTo()'` Cancel buttons with the CSP-safe `cactiReturnTo` class so the pages no longer trip Cacti's Content-Security-Policy script-src-attr directive
 * security: Move the UPS filter selects' inline onChange handlers into the ready block so the page no longer trips Cacti's Content-Security-Policy script-src-attr directive
 * dev: Enforce patch coverage of changed lines in CI and remove the inert COMPOSER_ROOT_VERSION env from the Pest step

diff --git a/INFO b/INFO
@@ -1,6 +1,6 @@
 [info]
 name = apcupsd
-version = 1.3.1
+version = 1.4.0
 longname = UPS Monitor Plugin for Cacti
 author = The Cacti Group
 email = 

diff --git a/README.md b/README.md
@@ -1,11 +1,13 @@
 # apcupsd
 
-This plugin is to be used to track UPS status from the `apcupsd` daemon as 
-well as SNMP enabled UPS'.  The `apcupsd` daemon is able to provide information 
-on APC and other types of UPS' over various network typologies such as USB, 
-Modbus, Ethernet, etc.  I have added basic SNMP support in order to provide
-direct access to certain models over the SNMP protocol thus removing the need
-for an `apcupsd` host to provide the monitoring details as well.
+This plugin is to be used to track UPS status from the `apcupsd` daemon, from
+the Network UPS Tools (NUT) `upsd` server, as well as from SNMP enabled UPS'.
+The `apcupsd` daemon is able to provide information on APC and other types of
+UPS' over various network typologies such as USB, Modbus, Ethernet, etc.  NUT
+provides the same for a very wide range of UPS vendors.  I have added basic
+SNMP support in order to provide direct access to certain models over the SNMP
+protocol thus removing the need for an `apcupsd` host to provide the monitoring
+details as well.
 
 Each `apcupsd` UPS is defined by a network address and port name.  Cacti uses 
 the `apcaccess` command to gather statistics from the UPS and store them 
@@ -16,14 +18,19 @@ Cacti database can be used for Reporting, Graphing, and Alerting.
 
 This plugin allows Cacti Administrators to easily track their Data Center
 UPS's that do not support SNMP protocol, but that are supported by the
-APCUPSD daemon.
+APCUPSD daemon or by NUT.
 
 ## Features
 
 * Track UPS status'
 
 * Log information to Cacti database
 
+* Three collector types: `apcupsd` (via `apcaccess`), SNMP, and NUT (via the
+  native NUT network protocol)
+
+* Fleet and per-UPS dashboards, with sortable tables
+
 ## Cacti compatibility
 
 If you are running a version of Cacti below 1.2.31, please add the function
@@ -41,25 +48,89 @@ $(document).off('click.cactiReturnTo', '.cactiReturnTo')
 ## Installation
 
 Install just like any other plugin, just throw it in the plugin directory, and
-Install and Enabled from the Plugin Management Interface.  Your Cacti install
-will require the Linux `apcupsd` package be installed and `apcaccess` be in the
-system path.
+Install and Enabled from the Plugin Management Interface.  To monitor
+`apcupsd` UPS', your Cacti install will require the Linux `apcupsd` package be
+installed and `apcaccess` be in the system path.  NUT UPS' need no extra
+software on the Cacti poller.
 
 After you have installed the plugin, install the Device Package in the 
 templates folder from Console > Import/Export > Import Package.  This 
 Device Template will be used for showing available graphs.
 
+## Monitoring UPS' with NUT
+
+Support for Network UPS Tools (NUT) is new in version 1.4.0.  The plugin talks
+to the NUT `upsd` data server directly with a native PHP implementation of the
+NUT network protocol, so the NUT client tools (`upsc`) do not have to be
+installed on the Cacti poller, and `curl` is not involved.  The only
+requirement is that the poller can open a TCP connection to `upsd`, which
+listens on port 3493 by default.
+
+To add a NUT UPS, go to the UPSes page, add a UPS, and choose the `NUT Based`
+type:
+
+* **Hostname** and **TCP Port**: the host running `upsd` and its port (3493).
+
+* **NUT UPS Name**: the name of the UPS on that server, as listed by
+  `upsc -l <host>` (the `ups` in `ups@host`).  One `upsd` can serve several
+  UPS', so add one UPS here for each of them.
+
+* **NUT Username** and **NUT Password**: optional.  NUT servers have always
+  allowed anonymous read-only access and that is how they are normally
+  monitored, so leave these blank unless your `upsd.users` is set up to require
+  authentication.  A user with no actions is enough, as the plugin only reads
+  data.  The password is stored in the Cacti database the same way SNMP
+  community strings and passwords are.
+
+* **Use TLS (STARTTLS)**: encrypts the connection before any credentials are
+  sent (the `openssl` PHP extension is required).  The server certificate and
+  host name are verified, so by default the certificate must be trusted by the
+  Cacti poller.  For a homelab with a self-signed certificate there are two
+  ways around that:
+
+  * **CA / Certificate File** (recommended): the full path, on the poller, to a
+    PEM file holding the CA certificate, or the self-signed server certificate
+    itself (for example `/etc/nut/upsd.pem` copied from the NUT server).  The
+    server is still authenticated against that file, and the Hostname must
+    match a name in the certificate.
+
+  * **Allow Self-Signed Certificate**: accepts any certificate without
+    checking it.  The connection is encrypted but the server's identity is not
+    verified, so a man-in-the-middle could still read the NUT credentials.  Use
+    it only on a network you trust, and prefer the certificate file.
+
+* **Timeout (seconds)**: how long the poller waits for the NUT server, from 1
+  to 120 seconds (5 by default).  It bounds the connection attempt and each
+  reply, and the connection plus the whole conversation together never take
+  longer than twice the timeout, so an unresponsive server cannot hold up the
+  poller.  The same setting applies to `apcupsd` UPS', where `apcaccess` is
+  stopped if it has not finished in time.  SNMP UPS' keep using the SNMP
+  timeout.
+
+On the NUT server, `upsd` only listens on localhost by default.  To let a
+Cacti poller reach it, add a `LISTEN` line to `upsd.conf` (for example
+`LISTEN 0.0.0.0 3493`), restart `upsd`, and allow the poller through the
+firewall.
+
+NUT reports power state with short flags.  These are translated into the
+`apcupsd` vocabulary when the data is saved (`OL` becomes `ONLINE`, `OB` becomes
+`ONBATT`, `LB` becomes `LOWBATT`, `RB` becomes `REPLACEBATT`, and `OVER`
+becomes `OVERLOAD`), so the dashboards and fleet totals treat NUT and `apcupsd`
+UPS' the same way.  A Cacti device is created for each NUT UPS automatically,
+the same as for `apcupsd` UPS'.
+
 ## Possible Bugs
 
 If you figure out this problem, see the Cacti forums!
 
 ## Future Changes
 
 Currently the SNMP based collection is functional but only covering APC UPS'.  
-I will likely fix that in the next release.  Need some help from others.l
+I will likely fix that in the next release.  Need some help from others.
 
-I've been looking at the NUT product to see if it's something I would add
-support for down the line, but as of yet, I have not made that decision.
+Measured real power (`ups.realpower`) is reported by many NUT UPS' but is not
+stored yet; the plugin continues to estimate watts from the nominal power
+rating and the load.
 
 Got any ideas or complaints, please log an issue on GitHub.
 

diff --git a/css/apcupsd.css b/css/apcupsd.css
@@ -524,36 +524,33 @@
 	white-space: nowrap;
 }
 
-/* Sortable column headers: the button inherits the header look and shows an
-   arrow for the active direction (exposed to assistive tech via aria-sort). */
-.apcupsdDashSortBtn {
-	appearance: none;
-	background: none;
-	border: 0;
-	padding: 0;
-	margin: 0;
-	font: inherit;
-	font-weight: inherit;
-	color: inherit;
-	cursor: pointer;
-	text-align: inherit;
+.apcupsdDashTable td.apcupsdDashNum,
+.apcupsdDashTable th.apcupsdDashNum {
+	text-align: right;
+	white-space: nowrap;
 }
 
-.apcupsdDashSortBtn:hover,
-.apcupsdDashSortBtn:focus-visible {
-	text-decoration: underline;
+/* Sortable column headers (same behaviour as the HMIB dashboards): click or
+   press Enter/Space to sort; the arrow shows the active direction. */
+.apcupsdDashTable th.apcupsdDashSortable {
+	cursor: pointer;
+	user-select: none;
 }
 
-.apcupsdDashSortable[aria-sort="ascending"] .apcupsdDashSortBtn::after {
-	content: " \25B2";
-	font-size: 0.75em;
+.apcupsdDashTable th.apcupsdDashSortable:hover {
+	background: color-mix(in srgb, currentColor 8%, transparent);
 }
 
-.apcupsdDashSortable[aria-sort="descending"] .apcupsdDashSortBtn::after {
-	content: " \25BC";
-	font-size: 0.75em;
+.apcupsdDashTable th.apcupsdDashSortable::after {
+	content: '\2195';
+	margin-left: 5px;
+	opacity: 0.4;
+	font-size: 10px;
 }
 
+.apcupsdDashTable th.apcupsdDashSortAsc::after  { content: '\2191'; opacity: 0.9; }
+.apcupsdDashTable th.apcupsdDashSortDesc::after { content: '\2193'; opacity: 0.9; }
+
 /* Fleet total row under the UPSes table. */
 .apcupsdDashTotalRow > td {
 	border-top: 2px solid color-mix(in srgb, currentColor 22%, transparent);