TarleyNode 0.4.2
Released 11 November 2026
Added
- FreeSWITCH and OpenSIPS are monitored built in: choose either as the Protocol in Add device (search for sip or voip) or assign its login on a device's page. FreeSWITCH is read through its event socket with its password: calls, sessions and their rate against the limits, registrations, uptime and each SIP gateway (one down fails the check, by name; one not registered warns). OpenSIPS is read through its management interface over HTTP (mi_http), with optional basic authentication and HTTPS (without HTTPS a password crosses the network in clear text): active dialogs, registered users and contacts, transactions, shared memory use and each dispatcher and load balancer destination (one probing, or a set or group with none active, fails; one switched off warns). A check can watch one gateway or destination by name. FreeSWITCH with mod_snmp is also read over SNMP without importing a MIB. See docs/voip.md.
- Each device has its own limit on how many readings automatic discovery adds: Edit device › Readings added automatically, up to 40 as before, 100, 200, 500 or no limit (up to 512 readings a device, each one polled; any number up to 10000 through the API), for core switches with hundreds of ports and sensors. Every reading counts against the licence. Only administrators can change it, and each change is in the audit trail. Readings found from this update on that the limit left out are added at the next rescan after it is raised. Readings found before this update are not added by themselves, since they cannot be told apart from ones whose check someone deleted: after raising the limit the console says how many readings are not monitored and offers to add up to the room under the new limit, the most important first, and the Sensors list offers the same and has a Not monitored filter to pick them one by one. A reading whose check someone deleted, even one ticked by hand, is never added back by discovery. When the licence has room for only some new readings, discovery and Monitor add the most important ones that fit instead of none, and the rescan result or the Sensors list says how many the full licence left out. Automatic port monitoring has no limit of its own and is unchanged.
- FortiGate firewalls are read in more depth, with no MIB to import: besides CPU, memory and sessions, disk use, low memory (the shortage behind conserve mode), new sessions a second, the HA mode, each HA member's sync state (a member out of sync warns) and each IPsec tunnel by name (a tunnel down fails). Uptime and interface traffic come from the standard MIBs, as before. A tunnel or other named row that is renumbered to another object fails as moved instead of reporting the other one.
- FortiGate firewalls also report, where the unit has them: each SD-WAN health check's members by name and interface (state, latency, jitter and packet loss; a dead member fails), CPU, memory and sessions per VDOM when there are several, hardware sensors by name, as temperatures (°C), fans (rpm), voltages (V) or power supplies when the name says so (one in alarm fails), the age in days of the antivirus and IPS signatures, and each FortiAP by serial number (offline fails). Tables a unit lacks add nothing. A vendor reading that returns nothing now shows as not supported instead of passing with no value.
- More firewalls and switches are read in depth, with no MIB to import, and the new readings are added automatically on the next rescan (within each device's limit on readings added automatically, health first). Cisco IOS, IOS-XE, NX-OS and ASA: memory pools by name (64-bit counters where the device has them; pools full by design are left out), system memory, stack members (from two; a version mismatch, provisioned-but-missing or removed member fails), active IPsec and IKE tunnels, and the Cisco entity sensors (temperatures, voltages, fans, and optic light levels in dBm, which are listed to tick) that Nexus and newer Catalyst switches report instead of ENVMON. Cisco ASA also: each failover unit (one down fails; the "this device" mark moving after a failover keeps each check on its unit) and AnyConnect sessions. Juniper Junos (beta): each component's state (one that goes down fails), temperature, CPU and memory per routing engine and card, and the red and yellow chassis alarms. Palo Alto: HA peer state, TCP, UDP and ICMP sessions and GlobalProtect gateway use. pfSense (beta): pf running and its state table, as a count and as a share of its limit. Sophos: swap use. Any net-snmp agent (Linux and BSD servers and appliances): the 5-minute load average. Per-optic Cisco readings are listed, not added automatically, as on other switches.
- Reports › Operations reports can cover a set of servers instead of a whole site: choose a device category, tick devices, or both, together with the site. The incident table, its CSV, the PDF report (including its top 10 lists) and scheduled reports follow the choice, the PDF and scheduled summary say what they cover (such as "Category: Servers · 5 chosen devices"), and a schedule keeps its devices and category. Removed devices drop out of a schedule; removing a category takes it off the schedules that used it and pauses them, so they do not start reporting every device.
- Devices › Locations keep the order you give them, the same for everyone: drag a location above or below another (a line shows where) to put it there, among its siblings or in another parent, or choose it and use Move up and Move down. Dropping onto a location still puts it inside. A new location goes last instead of in name order, and existing locations keep the order they had. Only administrators can, as with other location changes, and each move is in the audit log. Location pickers name each location by its path, such as Cloud › AWS.
- Settings › Shift roster: phone alerts on a weekly rota. Shifts are whole weeks from Monday 07:00 to the next Monday 07:00 in the roster's time zone. People on the roster claim and give back their own weeks and see who has the others; administrators choose who is on the roster and assign or swap weeks. Whoever has the current week has phone alerts switched on and the rest of the roster has them switched off; in a week nobody has, or whose holder has been deactivated, everyone on the roster gets them. Their own alerts switch is replaced by "On shift until Mon 07:00" or "Off shift" in the console and the app. People not on the roster, administrators included, keep their own setting. Every claim, give-back, assignment, membership change and automatic switch is in the audit log. See docs/mobile.md.
- The phone app has an Overview tab, after Alerts: the console's health ring with healthy, down and stale counts, the three newest open alerts and, for people who see every device, the Top 10 lists and the disks and links filling up. Rows open the phone's alert and check screens, the counts refresh every minute, and refreshing keeps the chosen period and list.
Changed
- Update the server first, then each connected site. Connected sites need this version to take the new FortiGate, vendor and VoIP readings, to retry SNMP requests, and to report a v1 trap's own agent address, so that traps from a shared address reach the right device.
- Several devices can share an address in a site, such as a firewall on SNMP and a server on SSH port 2222 behind one public address (NAT, port forwarding). Add device says which devices already use the address and adds anyway, and a device can be moved onto a used address. Scans still add no second device at an address: a found address goes to the device monitored there over SNMP, else the oldest, and flows, neighbors and ARP entries from a shared address go to that same device. Syslog and traps go to the device they name, by the syslog host name or the trap's sysName against device names and learned sysNames, or by a v1 trap's agent address, and to that device otherwise. A refused device now says which field is wrong: the site or where it is monitored from.
- The phone app's tabs are Alerts, Overview, Devices and Settings. The Sites screen and its site maintenance buttons are gone from the phone; put a site in or out of maintenance from the console or TarleyNode Desktop.
- The phone app's check screen opens its graph on the check's first percentage reading (CPU, memory used, the fullest or named file system, port utilisation) instead of Availability, which stays in the list. Your own choice stays across refreshes, and each check starts from its own default.
- IP register: clicking a range opens its addresses in place, under its row, with the same counts, search, devices, badges and assign, reserve and free as the old dialog. Click again to close it; several ranges can be open at once. The assign and reserve form also opens under the address, and Escape closes it.
Fixed
- The Palo Alto high availability, Check Point policy and Check Point cluster sensors report their state again. Before, every result they gave was discarded.
- Settings › MIB library imports a vendor's MIB set by itself. The standard modules vendor MIBs build on (SNMPv2-SMI, SNMPv2-TC, SNMPv2-CONF, SNMPv2-MIB, IF-MIB, IANAifType-MIB, INET-ADDRESS-MIB, IPV6-TC, HCNUM-TC and SNMP-FRAMEWORK-MIB), and FORTINET-CORE-MIB so the FortiGate MIB imports by itself, are built in, as their OID assignments, and cannot be removed; importing the real FORTINET-CORE-MIB replaces its copy and brings the names the copy lacks. Files chosen together are imported in one go and resolve against each other in any order, so a product MIB no longer fails because its vendor's core module came later in the list. A module that still cannot resolve now names only the modules it really lacks, and modules imported earlier gain the names a later import, or this update, makes resolvable.
- Turning on notifications in the phone app now also turns on alerts to that person's phones, unless they turned them off in the console, so the app no longer says "On" while no alert is sent. Phones that already had notifications on get alerts turned on when the server starts. When alerts to phones are off, the app's Settings says so and offers Turn on alerts. Phones of people in an access group are left as they were.
- A device with incident history can be removed again. Removing it could fail with "in use" on some installations; its incidents are kept, closed, as before.
- SNMP checks no longer fail on a single lost reply: each request is now tried again once within the check's timeout.
- Devices › Rescan all shown no longer leaves the console blank. A page that cannot be shown now says "Something went wrong showing this page", with the error and Reload, instead of a blank screen, and moving to another page recovers.
- Monitoring a port with no name or description names its check by the port's index, such as Port 3, instead of "Port ".
- Table headings in reports, SLA and other console tables line up with their values, and number columns, top traffic and the site health widget align their figures, with thousands separators. Link forecast speeds read correctly in Arabic, in the console and the phone app.
Security
- People in an access group no longer get phone or browser alerts about devices outside their groups, or about incidents with no device (a business service or connected site), which the console does not show them either. Each skipped alert shows on the incident as cancelled because the person cannot see the device, not as a failed delivery. Shared destinations such as email, Slack and webhooks still get every alert, and administrators and people in no group get all, as before, except that a phone an administrator paired with a lower role follows the administrator's access groups, as the phone app already does.