Installation
sudo install -m 755 elx-healthmonitor-linux-amd64 /usr/local/bin/elx-healthmonitor
sudo /usr/local/bin/elx-healthmonitor --install-service
sudo systemctl enable --now elx-healthmonitor
# web interface: http://<server>:8555sudo apt-get install -y smartmontools| Path | What it is |
|---|---|
/usr/local/bin/elx-healthmonitor | The executable |
/etc/elx-healthmonitor/config.json | Settings, mode 0600 |
/var/lib/elx-healthmonitor/history.db | Metric history and the event log |
/etc/systemd/system/elx-healthmonitor.service | The unit |
--config <path> configuration file
--listen <address> web interface address, e.g. :8555
--log-level <level> debug | info | warn | error
--version version and build timeChoosing what to collect
The Settings section decides what is polled, how often and what is excluded; the Alerts section decides when you hear about it. Changes apply immediately, without a restart. Editing config.json and restarting the service has the same effect.
Alerts
A rule is a condition, a threshold and a hold-down. The hold-down is the part that matters: without it, a monitoring system reports every momentary spike, and within a week nobody reads its mail.
| Watch for | A threshold that works |
|---|---|
| Filesystem filling up | 85 per cent — enough warning to act without hurrying |
| Processor load | Above 90 per cent held for ten minutes, not ten seconds |
| Disk temperature | Above 50 °C sustained; a brief peak during a backup is normal |
| SMART | Any change in reallocated or pending sectors — no hold-down, report at once |
| RAID | Array degraded, or a non-zero mismatch count — always |
| Service | A unit that should be running and is not |
Mail supports three encryption modes: SMTPS, where TLS starts with the connection, usually on port 465; STARTTLS, where it is negotiated, usually on 587; and unencrypted, which belongs only on a mail relay inside your own network.
History
- raw_retention_days — how long readings are kept exactly as measured; three days by default.
- rollup_interval_sec — the bucket size after that; five minutes, storing the average, minimum and maximum of each bucket.
- retention_days — when everything is deleted; a month by default.
Thinning runs in six-hour windows so that it never holds the database long enough to be noticed.
When something is wrong
| Symptom | Where to look |
|---|---|
| The interface does not open | systemctl status elx-healthmonitor, and whether the port is already taken |
| The SMART section is empty | smartmontools is not installed, or the disks are behind a controller that does not pass SMART through |
| No temperatures at all | A virtual machine — there are no physical sensors to read |
| Mail is not delivered | The wrong encryption mode for the port: 465 is SMTPS, 587 is STARTTLS |
| Alerts arrive constantly | The hold-down is too short, or the threshold is inside the machine's normal range |
| History stops after a few days | Retention is set lower than you think, or the disk holding the database is full |
Frequently asked questions
Where are the agent's own logs?
journalctl -u elx-healthmonitor -f.Should the interface be exposed to the internet?
How do I remove the power buttons?
server.allow_power_control in the configuration, and reboot and shutdown disappear from the header.