andreas404/insightd-agent

•Updated 4 months ago

Lightweight monitoring agent for insightd — collects Docker and host metrics

Image

4.1K

andreas404/insightd-agent repository overview

⁠Insightd Agent

Lightweight monitoring agent for insightd⁠. Deploy on each host or k8s node to collect container and system metrics, reporting back to the central hub via MQTT.

Supports two runtimes:

Docker — auto-detected via the docker socket (default)
Kubernetes — runs as a DaemonSet, one pod per node, uses kubelet stats

⁠Quick Start (Docker)

docker run -d \
  --name insightd-agent \
  --restart unless-stopped \
  -v /var/run/docker.sock:/var/run/docker.sock \
  -v /:/host:ro \
  -e INSIGHTD_HOST_ID=my-server \
  -e INSIGHTD_MQTT_URL=mqtt://your-hub-ip:1883 \
  -e INSIGHTD_MQTT_USER=insightd \
  -e INSIGHTD_MQTT_PASS=your-mqtt-password \
  andreas404/insightd-agent:latest

The agent will immediately start collecting metrics and publishing to the hub.

Tip: The hub's Add Agent page generates a ready-to-paste docker run command with your MQTT details pre-filled.

⁠Quick Start (Kubernetes / k3s)

Run as a DaemonSet — see agent/k8s/⁠ for the manifests and docs/kubernetes-setup.md⁠ for the full guide.

kubectl apply -f https://raw.githubusercontent.com/goldenproductions/insightd/main/agent/k8s/rbac.yaml
kubectl apply -f https://raw.githubusercontent.com/goldenproductions/insightd/main/agent/k8s/daemonset.yaml

Edit the DaemonSet first to set INSIGHTD_MQTT_URL (and optionally INSIGHTD_HOST_GROUP). K8s mode is read-only — container actions and image update checks aren't supported (those are managed by the cluster).

⁠What It Collects

Metric	Docker mode	Kubernetes mode
Containers	Status, CPU, RAM, restarts, network I/O, block I/O, health checks	Same — sourced from kubelet cAdvisor
Host CPU	`/proc/stat`	kubelet `/stats/summary`
Host memory	`/proc/meminfo`	kubelet `/stats/summary` + Node API capacity
Host uptime	`/proc/uptime`	Node `metadata.creationTimestamp`
Disks	Usage per mount, days-until-full forecast	Same
GPU	Utilization, memory, temperature (NVIDIA)	Not collected
Temperature	CPU/system sensor readings	Not collected
Disk I/O	Read/write throughput	Not collected
Network I/O	Per-interface bandwidth	Not collected
Load average	`/proc/loadavg`	Not collected (kernel concept that doesn't map to a single node)

Metrics are collected every 5 minutes (configurable) and published to MQTT. In k8s mode, the values that aren't collected appear as null rather than misleading numbers — /proc and /sys inside the agent pod reflect the underlying machine's kernel, not the node.

⁠Additional Capabilities

Container actions — start/stop/restart containers remotely from the hub UI (opt-in)
Log tailing — fetch container logs on demand from the hub UI
Remote updates — update the agent from the hub UI via MQTT (opt-in)
Image update detection — compares local images against Docker Hub registry

⁠Configuration

Variable	Default	Description
`INSIGHTD_HOST_ID`	`local`	Unique name for this host
`INSIGHTD_HOST_GROUP`	—	Optional logical group label (e.g. `production`, `basement`) for the Hosts page
`INSIGHTD_RUNTIME`	`auto`	Container runtime: `auto`, `docker`, or `kubernetes`
`INSIGHTD_MQTT_URL`	—	MQTT broker URL (required)
`INSIGHTD_MQTT_USER`	—	MQTT username
`INSIGHTD_MQTT_PASS`	—	MQTT password
`INSIGHTD_COLLECT_INTERVAL`	`5`	Collection interval (minutes)
`INSIGHTD_ALLOW_ACTIONS`	`false`	Enable container start/stop/restart (Docker mode)
`INSIGHTD_ALLOW_UPDATES`	`false`	Enable remote agent updates (Docker mode)
`INSIGHTD_DISK_WARN_THRESHOLD`	`85`	Disk warning threshold (%)
`TZ`	`UTC`	Timezone

⁠Volumes

Path	Purpose
`/var/run/docker.sock`	Docker socket (required)
`/host`	Host filesystem mount (read-only, for disk/host metrics)

⁠Resources

~15MB RAM typical usage
No local storage needed
Multi-arch: linux/amd64, linux/arm64

⁠Links

Tag summary

Recent tags

Content type

Image

Digest

sha256:829a562c1…

Size

92.7 MB

Last updated

4 months ago

docker pull andreas404/insightd-agent