Skip to content
TarleyNode
How it worksFeaturesSecurityLicensingDownload
Request a demo

Founding testers wanted. TarleyNode is new, and I’m looking for 20 IT and network admins to try it and tell me what’s broken or missing. In return you get a full licence free for 12 months. Become a founding tester →

TarleyNode 0.3.9

Released 5 November 2026

Added

  • gNMI telemetry (beta; not yet tested on real routers) from routers and switches (Cisco IOS XR and NX-OS, Juniper, Arista EOS, Nokia SR OS and SR Linux): a gNMI login on a router, switch or firewall reads OpenConfig port traffic, errors and status, CPU, memory, optical power and BGP sessions as often as every 10 seconds. Paths a device does not offer are listed on the reading instead of failing it. It is never suggested: choose gNMI telemetry (beta) in Add device or under Show other login types on a device. Connected sites need this version to read it. See Getting started in the console.
  • Traffic breaks flows down by source and destination network (AS number and name) and by DSCP traffic class (CS0/BE, AF11-AF43, EF and so on). AS numbers come from the exporter, or from the free ip2asn table if you place it on the server or site. The AS and DSCP lists need both the server and the connected site on this version, so update the server first and connected sites after it. A connected site updated first keeps sending its source, destination and application lists until the server catches up. Without the table on the server, the Traffic page says where to put it.
  • Traffic keeps hourly summaries (top 20 per list) for a year as well as per-minute summaries for 7 days, so the Traffic page shows up to a year. Settings › Server › Traffic history changes both (1 to 31 days per minute, 7 to 1825 days hourly). The same card shows the space flow summaries take, an estimate for the days you type and, with the bundled database, the free disk, and warns when the estimate gets close to it.
  • Links filling up, on the Overview and each interface's page: when an interface's busiest 5% of traffic (each day's 95th percentile over 90 days) reaches 80% of its speed, at the rate it is growing. When forecasts cannot be read it says why, and it names the days of history used when there are fewer than 90.
  • Settings › People & sign-in can turn on sign-in with an Active Directory or OpenLDAP account (LDAP). People use their directory name and password in the normal sign-in form; their directory group decides their role. Connections are always encrypted (ldaps:// or StartTLS), and Test connection checks the settings and shows the groups and role a person would get. An option also counts membership through nested Active Directory groups. An account with a TarleyNode password keeps signing in with that password, and a directory email can never take it over. See Getting started in the console.
  • Settings › People & sign-in can also add a SAML 2.0 identity provider, such as Microsoft Entra ID, Okta, AD FS or Google Workspace. The sign-in page then shows "Sign in with single sign-on". Assertions must be signed, made out to this server and used once. A matching email can never take over an account that has a TarleyNode password. See Getting started in the console.
  • On licences that include custom branding, Settings › Server › Product name lets an administrator show their own name, such as Monitoring, instead of TarleyNode. Without that licence option the card is hidden and TarleyNode shows everywhere. The name replaces the logo at the top of the sidebar and on the sign-in page with plain text, and is used in the browser tab, the console's name when it is installed as an app and the console's on-screen text. The On call phone app follows too: its installed name, title, on-screen text and notifications that name no device. Licence, terms, privacy, open-source notices, About, updates and feedback to the TarleyNode team keep the TarleyNode name, as do emails, text messages and the app and notification icons. Leave it empty to go back to TarleyNode. Changes are recorded in the audit trail.
  • A GPS map, next to Network map in the sidebar, places devices by latitude and longitude in the colour of their state. A device without its own coordinates shows at its site's, and devices at one place share a marker in the worst state, which lists them. Set coordinates with Edit device, and with Add a site and Edit on the Discovery page. Access group members see only their own devices and do not get the coordinates of other sites. Map tiles come from OpenStreetMap unless an administrator changes the tile server or turns tiles off under Settings › Server › GPS map. The tile server sees the IP address of everyone who views the map, so you can use one you run yourself. See Getting started in the console.
  • MySQL and MariaDB checks also read threads running, queries per second, slow queries, failed connection attempts, the buffer pool hit ratio and the longest running query in seconds, leaving out replication and other server threads. The check's summary shows them. A reading the login may not take is left out and the check still passes. Connected sites need this version to read them. See Getting started in the console.
  • Database activity, a button for administrators on a MySQL or MariaDB check's page, shows the server's load at that moment: running queries longest first, the statements that took the most time in total, space by database and the largest tables. It is read once with the check's login, from wherever the device is monitored, and each part says what the login needs when it cannot read it. Query text is kept on the server for ten minutes, for the person who opened the view only, and never in the readings or history. Opening it is recorded in the audit trail. A connected site needs this version to answer it, and the view says so when it does not.
  • The phone app can be installed on Android. It shows Install app on the pair screen and in Settings › Notifications when Chrome offers it, says how to install from Chrome's menu when Chrome does not, and hides the button once installed. Chrome can now also offer to install the phone app and the console on a server behind a sign-in gateway.
  • Add device can set up a login in one step: choose a Protocol (SNMP, SSH, Windows, VMware, Database, UniFi controller, Proxmox VE or gNMI, the last two in beta), use a saved login or create one, and its first check is added with the device. If the login cannot be assigned, no device is added. The default is None. A search box above the list finds a protocol by name, vendor or system, such as Linux or MySQL, and Enter picks the first match.
  • The SNMP set-up card has a Show button for administrators, which fills the commands with the saved community or the SNMP v3 user, protocols and passwords; Hide puts the placeholders back. Each Show is recorded in the audit trail, without the secret. Passwords with a backslash now stay as typed in the card's commands.

Changed

  • SNMP checks of a server also read the server's own uptime, each disk's used bytes and the number of CPU cores, and port checks read the link speed. They leave snap images, Docker and Podman storage, shm and loop mounts out of the disks, so a snap image that is always full no longer raises a Disk space alert. Those mounts stop being recorded; their existing history stays. Connected sites do this once they are updated.
  • Sensor lists for SNMP checks, in the console and the phone app, show readable lines: Memory with used and total, one Filesystem line per mount, CPU and load per core, Interface state, speed and traffic, and Uptime. Memory and Disk space rows show their own values instead of CPU and load. Graphs, the measurement list of a dashboard chart and CSV exports name a filesystem by its mount, such as Filesystem /home used, instead of disk 65.
  • Remove device moved from the page header to its own card at the bottom of the device page, for administrators with no sensor open, so it is harder to press by accident. The dialog that asks for the device's name is unchanged.
  • The back link on a device page returns to the list view you opened it from and names it, such as a location or No location. The device list keeps its search, site, category and location in the page address, so a reload or the browser's Back button brings them back. Opened from elsewhere, the link still says All devices.
  • The optional box when starting a free trial now says what it covers: news, offers and a reminder if your server stops checking in. Starting a trial also sends the language the console is in and, when the box is ticked, which wording was agreed to.
  • A device's Logins card suggests logins from the device's own operating system before a scan's guess, no longer suggests UniFi by itself, and suggests Proxmox only on Linux servers and devices of unknown type.

Security

  • Builds use Go 1.27.3 and golang.org/x/net 0.62.0, which fix security vulnerabilities in their earlier versions.

Fixed

  • The backup made before an update can no longer be interrupted by a cluster server restarting the application part way through it.
  • Linux servers that got the Linux server (SNMP) template before 0.3.7 no longer read their CPU, memory and disks twice. On upgrade, the System resources check found by discovery takes the place of the template's CPU check, as it has since 0.3.7: it keeps its full history and takes the CPU name and alert rule, and open incidents, business services and dashboards that used the CPU check move to it. The duplicate CPU check is removed, which frees a sensor; its history only repeats System resources since the template was applied. A System resources check with its own alert rule, a CPU check you edited, and paused checks are left as they are. The audit trail lists each server changed.
  • Overview › Top 10 names what each row on Highest CPU, Highest memory and Fullest disks measures (CPU, Memory or Disk), instead of whichever check it came from, such as System resources. A device read more than one way says how, for example CPU (SNMP) and CPU (SSH). Scheduled reports show the same.
  • A device a scan could only name "Device" and its address takes the name its SNMP agent reports (for example web-01), the first time it answers, unless that name is only an address or localhost. A name you gave a device is never changed, and the audit trail records the rename.
  • Memory for Linux servers monitored over SNMP no longer counts the cache the system frees on demand, so it matches the server's own figure. Open memory alerts on such servers close after the update, as the reading drops to the real figure. Existing history keeps the old values.
  • Disks for Linux servers monitored over SNMP leave out the system's memory-backed mounts (/run, /run/user, /dev/shm and the like), as the SSH check already does, so a full /run/user no longer shows as a full disk. Disk space alerts caused only by those mounts close after the update. Those mounts stop being recorded; their existing history stays.
  • Saving a site's logins (Discovery › Logins) no longer clears them. The form could open before the saved logins had loaded and show None, and pressing Save then cleared the site's SNMP, SSH and Windows logins. It now waits for them. If a site lost its logins this way, choose them again.
  • Removed sensors stay removed. Editing a device, even only renaming it, applied its automatic templates again and brought back checks you had deleted, such as Ping. An edit now applies them only when the device's type or operating system changes, and automatic templates leave out checks you removed; a removed Ping also stays off when a template is applied later. Applying a template by hand still adds them back.
  • Dashboard chart widgets find their check again after a reinstall or restore, instead of only saying Check not found. A chart keeps its check's name and its device's name and address, finds the check of the same name on that device and saves the new link when the viewer may change the dashboard. Charts saved earlier get these names on upgrade, if their check still exists. A chart with no match says the check no longer exists and can be removed in the editor.
  • Assigning a Database login on a device page to a device with no database checks failed with a message about VMware. It now asks for the database type, port and database, and adds the check.
  • MySQL and MariaDB checks read connections without performance_schema, which MariaDB ships turned off, so the reading now appears there. The database field says it may stay empty for MySQL.
  • A scan's device type follows a network maker named in the SNMP answer rather than the network card's maker, which can belong to another machine, so a wireless access point seen with a virtual machine host's card address is no longer typed as a Linux server.
  • The phone app's Check page fits the screen: a long measurement name no longer pushes the chart, the period buttons and the time axis past the edge.

All releases · Plain text

TarleyNode Self-hosted network monitoring.

Sales: sales@tarleynode.com. Support: support@tarleynode.com

How it worksFeaturesSecurityLicensingDownloadAboutComparisonsGuidesReleasesSupportContactPrivacyCookiesTermsRefunds
YouTubeGitHubXBlueskyFacebookInstagramThreads