What’s new in TarleyNode 0.4.3.
Released 16 November 2026. Updates install from the console whenever an administrator chooses.
Highlights
- Monitor Linux servers over SSH: services, time sync, processes and scheduled jobs, with a ready-made template.
- Process checks keep up with very busy servers.
- A clearer shift roster that keeps every past week as an on-call record.
- API keys are now read only and for administrators only.
- Find an address or pick a spot on a map when placing sites and devices.
Before you update
- Update the server first, then each connected site.
- API keys are now read only. Scripts that use a key to add, change or delete will be refused.
- Keys made by operators and viewers keep working for reads. Only administrators can now see and revoke them.
- The map now always uses OpenStreetMap. A tile server you set, or tiles turned off, no longer applies.
- Sending OpenTelemetry metrics with an API key is paused for now. See docs/opentelemetry.md.
New
- Linux health (SSH) reads systemd services, time sync, TCP, threads, the mail queue and read-only mounts.
- Linux process (SSH) check, and CPU and memory in the SNMP process check.
- Linux scheduled jobs (SSH) follows cron jobs run through a small wrapper. See docs/scheduled-jobs.md.
- A built-in Linux server (SSH) template, and SSH checks that use each device's own login.
- Find an address or pick on a map to set a site's or device's location.
- A Report button on the device page makes a report for that device only.
- Past shift roster weeks are kept as a permanent record. Administrators can download the history as CSV.
- An Every route tab on the API keys card lists each route and whether a key may use it.
- A how-to for the MIB library, and updated how-to guides and training manual.
Improved
- The phone app opens a check's graph on its main reading, such as CPU, certificate days left or port traffic.
- When a database check's login cannot read replication, its page now names the privilege to grant.
- What's new and the update prompt are easier to scan, with highlights first.
- Settings is rearranged: phones and Shift roster under People & sign-in, API keys under Server, MIB library under Network.
- Shift roster shows weeks as a table with badges, and the people on it in their own card.
- Location tree counts now match the colour of the location's name, and turn blue when selected.
- Rescan for changes is now in the device page header, beside Edit device.
- The SNMP set-up card shows only until the device's SNMP answers.
- The Top 10 card opens on Highest CPU. Least available stays in emailed and scheduled reports.
- After an update, browsers load the new console straight away instead of an old copy.
- The Edit check OID field says it takes MIB names too.
- The map always uses OpenStreetMap tiles and Nominatim address search. The privacy notice names both.
Fixed
- Devices with more than 500 checks list all of them, also when picking checks for services and dashboards.
- A check with no name no longer blanks the phone app.
- An SLA at 99.8996% no longer counts as meeting a 99.9% target.
- Health, availability and report figures no longer round up to 100% when something was down.
- One sensor a device cannot provide no longer marks the whole device Unsupported.
- Process checks no longer count zombie processes as running, and find process names longer than 15 characters.
- Check results no longer stall, and alerts open again, when an incident was closed some other way.
- Add check no longer keeps the old protocol's port, such as 443, after you switch to SNMP.
- An upgrade with under 10 GB of free disk now stops before installing anything, not half-way through.
- A device's report lists only its own site. Top lists in reports for chosen devices rank only those devices.
- Long device names and card titles are no longer squeezed by buttons beside them.
Security
- Grafana's Prometheus queries still work with read-only API keys.
- Only administrators can make, see and revoke API keys.
- API keys can no longer reach backups or device configuration backups.