​
Metrics Overview

The current server agent collects a focused set of host-level metrics and selected runtime signals, then sends them to Watchman Tower as JSON.

​
Static Host Information

On startup, the agent reports static platform information such as:
  • agent name
  • agent version
  • operating system
  • platform name
  • platform version
  • kernel version
  • architecture
This helps Watchman Tower associate telemetry with the running agent version and host environment.

​
Dynamic Metrics

The current dynamic payload includes:
  • cpu_usage
  • memory_usage
  • total_memory
  • used_memory
  • disk_usage
  • total_disk
  • used_disk
  • network_sent
  • network_received
  • load_1
  • load_5
  • load_15
  • uptime
  • swap_total
  • swap_used
  • swap_usage

​
Process Metrics

The agent also sends a sampled list of high-usage processes. For each process entry, the payload currently includes:
  • PID
  • process name
  • CPU usage
  • memory usage
The current implementation sorts processes by CPU usage, keeps a bounded top set, and may reuse cached process snapshots between main metric submissions to reduce host overhead.

​
Service and Port Health

The current agent can also send lightweight server health blocks such as:
  • service_health
  • port_health
  • docker_health
  • nginx_health
  • mongo_health
  • redis_health
These signals are intentionally lightweight and are meant to improve host and runtime visibility without turning the current agent into a full observability stack.

​
Nginx Discovery

When Nginx is available and its configuration is readable, the current agent can also report:
  • discovered server_name values
  • root paths when present
  • Nginx config file paths
  • listen ports for discovered virtual hosts
This data is exposed as nginx_sites and is useful for understanding what may be running on a given server.

​
How the Backend Uses These Metrics

The backend stores server stat entries and also keeps a current server snapshot for list and detail views. That allows Watchman Tower to surface:
  • current CPU, memory, and disk usage
  • recent server stat history
  • current agent version on the server record
  • process samples included in metric submissions
  • service, port, and runtime health surfaces
  • Nginx discovery metadata where available

​
What This Does Not Yet Mean

These metrics are useful today, but docs should stay honest about scope. The current implementation does not imply:
  • advanced observability pipelines
  • service-level tracing
  • automatic metric enrichment by Noema today
  • custom metrics ingestion from user-defined collectors
  • full infrastructure autodiscovery across arbitrary software stacks

​
Next Steps