Features01 Dashboard

The whole engine, at a glance.

Engine status, the counts of what you have, CPU, memory and network across all containers live and over time, where your disk space went, and one card with everything that needs attention. Open the app and you know where you stand.

Where do I look first?Read the comic: is the kitchen open, how busy is it, and what needs you →
Fig. — The dashboard: the counts in one row, the live charts under them, and the history in one panel with a tab per kind.

What you can do

Engine status

Is the daemon there, and which context are you on. Shown on the dashboard and in the status bar, with the version of the app at its end.

Why it does not start

When the engine does not answer, the app looks outside Docker: in WSL as root, on a server over SSH with the same login. It recognises a broken or unknown setting in daemon.json, -H fd:// next to hosts, live-restore on a swarm node, a stale pid file, a taken port, a full disk and more, and fixes what it recognised: a backup back, a key out, a drop-in, a restart. Without root it gives you the commands to run yourself.

CPU, memory and network

Three live charts side by side under the counts: CPU, memory and network across all running containers, so a runaway process shows up straight away. The numbers of right now sit in the row above them.

Further back

One panel with a tab for CPU, memory, network and disk. Switch it between 24 hours, 7 days, 30 days and 12 months, or pick a span of your own up to a year back. The further back, the coarser: a point per five minutes, per hour, per six hours, then per day. It is kept on disk, so it survives a restart. The same switch sits above the charts of a container. For a machine you follow, the charts include what the app measured there while you were elsewhere.

Long tasks in one list

The status bar lists what is running and what just finished: compose up and down, updates, backups, scans, pulls, pushes, builds and moving a project to another machine, with how long it took, the step it is in, and the reason when one failed. A pull, a push, a build, a rebuild of one service, compose up or down, updating now, a backup, a scan and a move to another machine can be cancelled there, and a compose watch stopped.

Docker engine settings

Edit the daemon.json of the engine in three tabs: a form with the common settings, the file itself as JSON in colour, and its earlier versions. The form has registry mirrors, proxy, log rotation, DNS and the cache limit of the builder; an empty field goes back to the default of Docker. It is checked as you type, the old file is kept next to it, and Save and reload makes Docker read it again. The app tells you which changes still need a restart of Docker, and can restart it for you: the host checks the file first, and the app waits until Docker is back. If it does not come back, the host itself puts the previous file back after two and a half minutes and starts Docker again, keeping the one that failed. Without systemd, the restart goes through the init script. The last five versions are kept: compare one with the editor line by line, give it a note, or load it back, each from its own menu. A change to who can reach Docker, or to which program it starts as root, asks for confirmation in a window of the app itself. On a machine in a swarm, live-restore stays off: Docker takes it on a reload, but no longer starts after the next boot.

Where the space is

Disk usage split into images, volumes, build cache and containers, with the ten largest of each and the way to remove them.

Needs attention

Crash loops, failing health checks, containers at their memory limit or without their network or ports, newer images, containers still on an older image, compose drift and newer or changed templates from the catalog, on one card, with what to do next to each line. Clear hides what you already know; a line comes back only once it was gone for a while and turns up again.

Before it goes wrong

Needs attention also looks ahead: memory that will reach its limit within a day, a container that falls over and restarts far more often than last week, one that uses far more CPU, network or disk I/O than in the days before, a disk that is full within a month, and a memory limit that turned out too tight. The same for a machine you follow, from its own measurements. A server that boots in the morning is not a restart.

Recently stopped

Which containers stopped, and whether they exited cleanly or with an error code.

Clean up per kind

Prune images, containers, volumes, networks or build cache from one dialog, with the count per kind. Named volumes without a container are a choice of their own, off by default, with their names in view. The networks of a swarm never count as unused.

Engine settings

A mistake in daemon.json is caught before it reaches the host.

A wrong value in this file can stop Docker from starting, and then the app itself loses its connection. So the file is checked as you type: a number where Docker wants text, a mirror without https, a setting it does not know. Save stays off until it is right. A helper container writes the file on the Docker host, next to a backup, with the owner and permissions of the old one; if the file changed on the host in the meantime, the app says so instead of overwriting it. Docker Desktop manages this in its own settings, so there the app leaves it alone.

Fig. — A daemon.json with an error; Save stays off until it is fixed.

Made for WSL

When the daemon doesn't answer, the app offers to start it.

WSL shuts down a distro without an open session after 15 seconds. If that happened, or the daemon inside the distro isn't running, the dashboard says which one it is and gives you the button to fix it.

All features

Try it on your own daemon.
Free and open source. Windows installer, Linux AppImage, or npx on any platform.
Download for Windows
$ npx docker-client-mx