Skip to main content

System Monitoring

Use the eEVOS monitoring functions to check individual systems and clusters during daily operation. The built-in dashboards provide a quick overview. Prometheus, the REST API, and Ansible support detailed queries and centralized monitoring of multiple installations.

Monitoring options at a glance

Method Best suited for Access
eEVOS dashboards Daily checks and charts Normal GUI login
Prometheus Metrics, capacity analysis, and PromQL https://<eevos-address>/prometheus/ after GUI login
REST API Structured health, inventory, and job data API key via https://<eevos-address>:18443/api/v1/
Ansible Repeatable checks across several systems Separate control node using SSH or the REST API
Alerts Proactive notification of service and capacity problems Configured notification channel

Use the built-in web monitoring

  1. Sign in to the eEVOS administration interface.
  2. Open Statistics.
  3. Select IO Stats, VM Statistics, or Ceph Stats.
  4. Choose a time range and compare all cluster nodes.
  5. Investigate sustained high values, clear deviations, missing series, and unexpected zero values.

eEVOS system dashboard

The system dashboard provides a quick view of health and utilization.

eEVOS performance monitoring

Area What to look for
CPU and load Sustained utilization, sudden changes, or one node carrying more load than the others.
Memory Low available memory, increasing use, or differences between nodes.
Storage and network High I/O wait, continuously saturated links, or inactive interfaces.
Virtual machines CPU, memory, block-I/O, and network behavior for individual VMs.
Ceph OSDs that are not up/in, degraded placement groups, recovery activity, latency, and capacity.

Use Prometheus

Prometheus collects time-series data from system, VM, and Ceph exporters. Open Prometheus through the eEVOS proxy. Do not expose the local Prometheus port directly to an untrusted network.

  1. Sign in to the eEVOS interface.
  2. In the same browser, open https://<eevos-address>/prometheus/.
  3. Use Query for metrics, Alerts for active alerts, and Status → Targets for exporter health.

Prometheus query in eEVOS

Useful Prometheus queries

Purpose PromQL
Exporter availability up
System load node_load1
Available memory percentage 100 * node_memory_MemAvailable_bytes / node_memory_MemTotal_bytes
Non-idle CPU usage 100 * (1 - avg by(instance) (rate(node_cpu_seconds_total{mode="idle"}[5m])))
Ceph OSD health ceph_osd_up
Used Ceph capacity ceph_cluster_used_bytes

Metric names can differ between editions and exporter versions. Use the metric search to confirm the names available on your system.

Configure the REST API for monitoring

  1. Open System Settings → API Configuration.
  2. Enable external REST API access.
  3. Select a reachable management network. Localhost only cannot be reached remotely.
  4. Use HTTPS and the default port 18443.
  5. Create a dedicated API user with a strong key and the minimum required permissions.
  6. Save the settings and test the health endpoint from the monitoring host.

eEVOS REST API settings

Basic API test

export EEVOS_URL="https://<eevos-address>:18443"
export EEVOS_API_KEY="<api-key>"
curl --fail --silent --show-error   -H "X-API-Key: $EEVOS_API_KEY"   "$EEVOS_URL/health"

Useful endpoints include /health, /cluster, /nodes, /vms, /storage, and /jobs. Confirm the endpoints available in your installed version before building permanent checks.

Monitor with Ansible

Use Ansible when the same health checks must run across several eEVOS systems. Keep the inventory and playbooks on a dedicated control node.

  1. Add the management addresses to an inventory group.
  2. Use SSH modules for operating-system checks or ansible.builtin.uri for REST API checks.
  3. Store API keys in Ansible Vault or another secrets manager.
  4. Fail the play when a required service, target, node, or storage component is unhealthy.
  5. Run the play manually, from a scheduler, or from your monitoring platform.

Alerts and check intervals

Check Suggested interval
Node, exporter, and API availability Every 1–5 minutes
Ceph health and storage faults Every 1–5 minutes
VM and job status Every 5–15 minutes
Capacity trend Hourly or daily

Troubleshooting

Problem Check and action
Prometheus redirects to the login page Sign in to the GUI first and open the proxy in the same browser.
A target is down Check Status → Targets, lastError, the node, exporter service, DNS, and firewall.
A dashboard contains gaps Check Prometheus readiness, target health, time synchronization, and service logs.
REST works only locally Check the listening network, firewall, routing, and external API access.
REST returns 401 or 403 Check the API key, role, expiry, and key rotation.
REST returns 429 Reduce the request frequency.

Daily checklist

  • All expected Prometheus targets report up.
  • All nodes provide current system data.
  • Ceph OSDs and placement groups are healthy.
  • No unexpected critical alerts remain open.
  • REST /health responds and credentials are not close to expiry.
  • Jobs do not remain in running state indefinitely.
  • VM placement and balancing match expectations.
  • Capacity leaves enough reserve for failures and maintenance.

Security

Use the GUI session cookie only for the Prometheus proxy and a dedicated API key for REST. Restrict access to management networks, validate TLS certificates, and rotate credentials regularly. Do not publish Prometheus, exporters, or the REST API directly to untrusted networks.

Query terminology

PromQL is the Prometheus Query Language. Use it to select, filter, and calculate time-series metrics in Prometheus-compatible dashboards and API queries.

Back to top