Dashboard

← Manual index

The landing view summarises the selected host.

Host facts

Top row of cards: hostname + Docker version, CPU count + architecture, total memory + OS, and counts of running / stopped containers and images.

Disk usage

A breakdown from docker system df:

• Layers total — combined size of image layers on disk.
• Images / Containers (rw) / Volumes / Build cache — count and reclaimable size per category.

Use this to spot bloat; reclaim space from Images (prune dangling), Volumes (prune unused) or the build cache.

Resource usage

Three panels: CPU and memory as a share of the host, and network as current throughput.

The two pie charts show how the running containers divide up the host's CPU and memory — i.e. what slice of the whole machine each container is using right now, with the unused remainder shown as Free. The busiest containers get their own slice; the rest are grouped as Other.

CPU share is relative to all cores (100% = the entire host); memory share is usage ÷ total RAM. Remote hosts work the same, over the Docker API.

Network · all containers is the host-wide RX/TX rate right now, with a short rolling trend beside it. It is deliberately not a pie or a top-talkers list: a pie claims "parts of a whole", and the only whole available is whatever happens to be moving — so one container at 100% of 2 KB/s would look exactly like one at 100% of 800 MB/s. A live ranking is no better, because throughput is bursty enough to reorder itself on every poll. Per-container series live on the container detail.

Summed across running containers, so container-to-container traffic counts twice — once as one side's TX and once as the other's RX.

All three panels re-sample on Docker lifecycle events and on a slow poll, updating in place; a transient error keeps the last good numbers rather than blanking the section.

Open ports

A host-wide map of every published port across the running containers. Scan needs write access to the dashboard section, because it is an active network action rather than a lookup — a read-only account can see the port map but not launch the probe. It actively connects to each one and fingerprints what's really listening (SSH / HTTP(S) / SMTP / Redis / TLS / banner) — not just a guess from the port number. It only runs on demand (probing is an active network action), works for remote hosts too (SSH ports are tunnelled), and only touches your own hosts.

Running containers

A live table (refreshes automatically) of what's running, with quick start/stop/restart actions. Click a name to open its detail page. “View all →” goes to the full Containers list.

Multi-host: the dashboard always reflects the host chosen in the sidebar switcher. Switch hosts to see another daemon's numbers.