Devpilot
Monitoring

Monitoring Overview

See what every server is doing right now and what it did over the last year — live metrics, processes, services, disks, network, availability, alerts, and the activity trail behind every change.

Devpilot's monitoring answers the questions you actually ask about a server: Is it up? What is it doing? What changed? Will it run out? Who should know? It is built on the Devpilot agent installed on each server and shows up in two places: a workspace-wide Fleet view and a per-server Monitoring area.

What you can monitor

How data gets there

Every connected server runs the Devpilot agent. It samples the server locally and sends the samples to Devpilot — nothing on your server is exposed inbound.

Devpilot agent (current)Legacy agent
Sample intervalevery 15 seconds, shipped in one-minute batchesevery 3 minutes
CollectedCPU (overall, per core, I/O wait, steal), memory and swap, every real filesystem with inode usage, disk throughput, network throughput and errors per interface, TCP connection states, the top processes by CPU and memory, systemd services, Docker containers, host facts (OS, kernel, pending updates, reboot required)CPU, memory, disk usage, network counters
Heartbeatevery 30 seconds, independent of sampleswith each sample
Offline resiliencesamples are queued on the server and delivered when connectivity returnssamples during an outage are lost

A server's Agent & host card on its monitoring page shows the agent version it is running. Reinstalling the agent from the server page updates it to the current one.

How long data is kept

ResolutionKept forUsed by
Raw samples3 daysshort ranges (up to a few hours)
5-minute averages, minimums, maximums and 95th percentiles90 daysranges up to a few weeks
Hourlyabout 13 monthslong ranges and year-over-year comparison

Charts choose the resolution automatically for the range you pick, so a 30-day view never has to load every raw sample.

Live updates

Monitoring pages stay connected to Devpilot. A new sample appears on the charts within a few seconds of the agent collecting it, the status strip flips the moment a server goes quiet or comes back, and a new incident shows up without a refresh. The Live pill at the top of a page tells you the connection is up; if it drops, the page falls back to refreshing every minute.

Availability, honestly measured

A server is online while its agent heartbeats. If the heartbeats stop, the server is marked quiet for a short grace period and then offline — that transition is recorded as an incident with a start time, an end time when it recovers, and a duration. Uptime percentages, the 90-day status strip and mean time to recovery are all derived from those recorded outages, not from a self-reported "I'm up" flag.

Permissions

Anyone who can view the workspace can see monitoring. Creating and editing alert rules, acknowledging or resolving incidents, and installing or reinstalling the agent require the ability to manage servers in the workspace.