Raylux Manual

Appearance

Version 2026.20

Running Nexus ​

Day-to-day operation of an existing install.

Starting and stopping ​

powershell
sc query RayluxNexus
sc stop  RayluxNexus
sc start RayluxNexus
bash
systemctl status raylux-nexus
sudo systemctl stop raylux-nexus
sudo systemctl start raylux-nexus

The service starts automatically with the machine. If you stop it deliberately it stays stopped — it is configured to restart after a crash, not after an administrator stops it.

An administrator can also shut Nexus down from the web interface. That counts as a deliberate stop, so it will not restart on its own.

Watching the log ​

powershell
Get-Content C:\ProgramData\Raylux\logs\raylux_nexus.log -Wait
bash
journalctl -u raylux-nexus -f

Administrators can also read the log in the browser, on Nexus's Logs page, without access to the server.

There is no Windows Event Log integration

Start-up failures show up as a service-specific exit code and in the log file above, not in Event Viewer.

Health checks ​

Two endpoints answer without a sign-in, and disclose nothing about the installation. Point your monitoring at them.

EndpointMeaningUse it for
/healthThe process is alive and answering.Liveness — "should this be restarted?"
/health/readyNexus has finished starting and can serve.Readiness — "should traffic go here?"
bash
curl http://nexus:8080/health          # {"status":"ok"}
curl http://nexus:8080/health/ready    # {"status":"ready"}

/health/ready returns 503 until start-up completes. A load balancer should gate on readiness, not liveness — restarting a Nexus instance that is merely still starting makes an outage longer.

Keeping an eye on it ​

The Performance page trends CPU, memory, threads and handles. Nexus also publishes them as [SYSTEM] tags, so you can trend and alarm on them from a screen like any other value:

TagWhat it tells you
[SYSTEM]ResourceHealthA single overall verdict.
[SYSTEM]MemoryTrendMbPerHourRise of the memory floor (MB/h) over the evaluated window; 0 until a window can be judged.
[SYSTEM]MemoryGrowthStatelearning (warm-up or settling after a tag import, no verdict), ok, or growing.
[SYSTEM]MemoryGrowthNote, MemoryFloorMb, MemoryBaselineMb, MemoryTrendFit, MemoryTrendWindowMinThe evidence behind the verdict.
[SYSTEM]ThreadCount, [SYSTEM]HandleCountLeaks that memory alone would not show.

Memory is judged on the lowest reading in each 10 minutes, so warm-up, a project import and allocator churn do not count as a leak. There is no verdict for the first hour after start or for 30 minutes after a large tag-count change; growth is raised only after three consecutive agreeing checks, and cleared only after the floor has been flat for an hour.

On a new deployment, watch these over the first week.

In this section ​