The tab that opens first. It answers "is this machine healthy right now?" and, when it is not, "what is using it up?"
What is on the page
CPU, load, memory, swap
Live figures, refreshed while you watch.
- CPU — how busy the processor is.
- Load — how many tasks are waiting. Read it against the core count shown
beside it:
2.10 on 4 coresis comfortable;7.90 on 2 coresis not. - Memory — RAM in use. Linux uses spare memory for caching, so 70–80% is normal and healthy.
- Swap — emergency memory on disk. Steady swap use means the server wants more RAM.
Uptime
How long since the machine was last restarted. An uptime of two years is not a badge of honour — it usually means kernel security updates were never applied. Check Packages.
Disks and inodes
Every filesystem with how full it is and how much is free. Inodes are counted separately: they are how many files a disk can hold, and a disk can run out of them with plenty of gigabytes left.
Warnings appear at 80% full (inodes at 90%), and a full disk becomes a problem at 95%.
Network
Current traffic in and out.
Top processes
The ten biggest consumers, sorted by CPU or by memory.
Processes that share a command line are shown as one row, with their CPU and memory added up. Forty PHP-FPM workers read as one php-fpm: pool www row marked ×40 instead of filling the list. Tick Show each process to list them one by one.
Stop asks a single process to stop tidily (SIGTERM) and is recorded in the Activity log. A grouped row has Show each instead, so you pick which process to stop.
Listening ports
What is accepting connections, on which port, and which program owns it. Useful for two questions: "is my app actually listening?" and "what is exposed that should not be?"
Logged in
Who is signed into the machine right now.
Failed units
Services that tried to run and gave up. Anything here is worth opening in Services.
The warnings
Plain sentences appear at the top when a threshold is crossed — "Disk / is 96% full", "Memory is 93% used". They use exactly the same thresholds as the health cards, so a server never reads "fine" on one screen and "warning" on the other.
What to do about what you see
| What you see | Likely meaning | Next step |
|---|---|---|
| One process at 100% CPU, constantly | Stuck or overloaded app | Logs for that app, then restart it |
| Memory near 100% and swap climbing | Not enough RAM, or a leak | Restart the app; consider more RAM or swap |
| Load high but CPU low | Waiting on disk or network | Ask the AI assistant to investigate |
| Disk climbing every day | Logs or backups accumulating | Logs → largest logs |
| A port open you do not recognise | Something you did not install | Users, then ask the assistant |
Not sure? Press Ask AI bot on the server card, or open a chat and ask "what is using the memory on this server?" — it can read all of the above and explain it.
Looking back, and setting it up from chat
This page shows the server right now. To see what happened earlier, install the monitoring agent (the Monitoring page in the sidebar). It keeps 7 days of CPU, memory, swap and disk, and which containers, PM2 apps and services were using them.
You don't have to open each server to set it up. Ask the assistant:
- "Set up monitoring on all my servers" — it lists them, asks you once, and installs the agent on each.
- "Why did web-1 go down last night?" — it reads the history first: memory and swap climbing, the minute the server stopped answering, and which container or app was growing.
- "Monitoring says web-2 is unreachable" — it checks the agent and tells you the fix, such as opening port 8443 to the panel only on AWS.
Every install, removal or firewall change shows a Yes / No card first.
