Removing Provelox

Provelox lives entirely inside Docker, so removing it means removing Docker objects. The image has two commands for it, detach and uninstall. Both run on a Swarm manager, show what they would do first, and change nothing until you add --yes. They only touch objects that carry Provelox's own label, so your own services, networks and volumes are never listed or removed, and they never leave the Swarm. Plain uninstall is the gentle one: it removes Provelox's own console and API and leaves your apps running. Taking the apps down too needs --remove-apps. This page covers the three cases people ask about, all tried on a test server (Docker 29 with Swarm).

Every command below is run from the image:

docker run --rm -v /var/run/docker.sock:/var/run/docker.sock ghcr.io/provelox/provelox:latest <command>

It is written pvx-remove <command> for short in the rest of this page. To type it that way, define a shortcut in your shell once:

# bash or zsh
pvx-remove() { docker run --rm -v /var/run/docker.sock:/var/run/docker.sock ghcr.io/provelox/provelox:latest "$@"; }

# fish
function pvx-remove; docker run --rm -v /var/run/docker.sock:/var/run/docker.sock ghcr.io/provelox/provelox:latest $argv; end

Use the same image version you run (for example ghcr.io/provelox/provelox:0.5.0) so the command matches the platform. The shortcut lasts for that terminal window only.

What Provelox owns

Kind	Names
Its own services	pvx-provelox (the API), pvx-console, pvx-traefik (the proxy)
Your apps	one Swarm service per app, pvx-<app>
Networks	provelox-net, and pvx-net-<project> for each project
Volumes	pvx-provelox-data (its database), pvx-traefik-certs (certificates), and pvx-<app>-… for each app's data
Secrets and configs	pvx-provelox-jwt, pvx-provelox-admin-password, pvx-console-session, and two Swarm configs

Nothing is installed outside Docker. The only change to the host is that install starts a Swarm when the host has none.

List what is there before you remove anything:

docker service ls
docker network ls
docker volume ls

Everything Provelox made starts with pvx- or provelox-. Anything else is yours and is not touched by the steps below.

Remove Provelox and leave the apps running

This is what uninstall does when you give it no other option, because it is the one that does not disturb your running apps. Use it when you want to stop running Provelox's own console and API while your apps carry on. Apps are ordinary Swarm services and the proxy reads their labels straight from Docker, so they can keep serving, but not by simply removing the two services.

pvx-remove uninstall            # shows the plan
pvx-remove uninstall --yes      # does it

It first runs detach, then removes only pvx-console and pvx-provelox. The proxy, the apps, the secrets and all data stay. On the test server an app kept answering, and was not restarted, after the console and API were gone.

Why detach is needed

Each app's route has an error-page step that points at a small service defined by the API's own labels. When the API goes, that service disappears, the proxy rejects every route that uses it, and the apps answer 404 page not found even though they are running. (The proxy's log says: the service "pvx_error_pages@swarm" does not exist.) detach takes that step out of each route first. It only changes labels, so nothing restarts, and domains, HTTPS redirects and other settings are left as they were.

pvx-remove detach on its own does just that, and is safe to run any time: the apps then no longer depend on the API for their routes. Installing Provelox again puts the step back.

Apps protected by Sign in with… ask the API about every request. detach leaves them alone and names them, because taking their protection out would make them public. They refuse visitors once the API is gone, so turn that protection off first (in the app's Networking tab) or accept that they stay down.

What you give up

• Managing apps: no console, deploys, upgrades, rollbacks, scaling, or changes to domains and variables. Use docker service update by hand.
• Provelox's error pages: the page for an unknown address and the custom 502 to 504 pages are gone. Visitors see Traefik's plain "Bad Gateway" and "404 page not found".
• Background work: usage history, image update checks and certificate diagnosis stop. Certificates still renew, because Traefik does that itself.
• The platform addresses (console.<root-domain>, api.<root-domain>) stop answering.

Remove everything, apps included

Taking the apps down too is a separate, explicit choice: add --remove-apps.

pvx-remove uninstall --remove-apps            # shows the plan
pvx-remove uninstall --remove-apps --yes      # does it

Your apps stop and answer nothing until you install Provelox again. See Doing this on a production server before running it on one.

This removes the apps, the console, the API and the proxy, their secrets and configs, and the networks. Volumes are kept: databases, uploads and certificates. Add --delete-data (only allowed with --remove-apps) to remove those too, together with the apps' secrets. That cannot be undone, so back up what you need first.

Before running it you may prefer to delete apps from the console (tick Also delete data for those with storage), which removes each app's service, volumes and network together.

The Swarm itself is never left. If Provelox started it and nothing else on the host uses it, docker swarm leave --force removes it, by hand. If you also run your own Traefik or other stacks on this Swarm, do not.

If something is left behind (Docker will not remove a network while a container still uses it), the command says what and exits with an error. Run it again a minute later.

Next to your own Traefik

Provelox brings its own Traefik (pvx-traefik), which publishes ports 80 and 443 on the host. If you run another Traefik, install Provelox on other ports (--http-port, --https-port, see Install). Removing Provelox only touches the pvx- and provelox- objects above, so your own Traefik, stacks and networks are not affected, unless you attached them to provelox-net.

Where things live, and what each command changes

What	Where it lives
Apps' settings, environment variables, domains, ports, deploy keys, projects' compose files, collections, access tokens, registry and sign-in settings, the admin password's hash, activity and usage history	Provelox's database, on the pvx-provelox-data volume. Variables are also in each running service's definition.
App secrets (Secrets in an app)	Only in Docker's secret store. Provelox keeps their names, never their values, and cannot read one back.
Certificates	The pvx-traefik-certs volume
Each app's own data	Its pvx-<app>-… volumes
The platform's own secrets (sign-in signing key, first admin password, console session key)	Docker secrets pvx-provelox-jwt, pvx-provelox-admin-password, pvx-console-session

Command	Apps	Environment variables	App secrets	Data volumes and certificates	Console, API, proxy
detach	Keep running, nothing restarts	Unchanged	Unchanged	Unchanged	Unchanged
uninstall	Keep running, nothing restarts	Unchanged (still in the running services)	Unchanged (still mounted in the running apps)	Unchanged	Console and API removed, proxy keeps running
uninstall --remove-apps	Stop and are removed (their settings stay in the database)	Kept in the database, applied again by install	Kept: Docker still holds them	Kept	All removed
uninstall --remove-apps --delete-data	Stop and are removed	Lost with the database	Deleted	Deleted	All removed

After a reinstall over a kept database, apps come back with their settings, variables, domains and secrets (matched by name), and certificates stay valid. Signed-in sessions are not kept: everyone, you included, signs in again with the admin password you had, because the signing key is made again.

Also gone with uninstall --remove-apps: the networks (made again by install), the platform's own secrets (made again) and the placeholder pages. Images pulled onto the server stay in Docker's cache. Anything that is not Provelox's, such as your own services, networks and volumes, is never touched.

A secret is deleted only by --delete-data, or if you leave the Swarm yourself. Leaving the Swarm removes Docker's secret store, so those values are gone: write them down first if you may need them.

Doing this on a production server

What happens to running apps, as seen on the test server (an app on a fixed tag, one with a volume, and one on a tag that was moved to a different image on purpose, the way latest moves):

Step	Result for the running apps
detach	Nothing restarts. Only routing labels change.
uninstall	Nothing restarts. The apps kept answering the whole time.
Then install again (same version)	Nothing restarts: the same containers kept running, and the moved tag was not pulled again, so they stayed on the image they were running.
uninstall --remove-apps (data kept)	Every app stops and the server answers nothing until you install again.
Then install again	Apps are created again from their saved settings and pull their tag again. An app on latest, or any tag that moved, can come back as a newer image. In the test, the app on the moved tag came back as a different image and failed.

So for production:

• Use plain uninstall. It is the default because it is the only form that leaves apps running.
• Use uninstall --remove-apps only in a maintenance window, and pin every app to a version tag (1.26-alpine, 20261030-013846) first, so what comes back is the image that was running. The command lists the apps that will stop and marks the ones on a tag that can move.
• Install again with the same Provelox version if you want nothing to restart. A newer Provelox applies every app again when it starts, and Docker restarts any app whose definition that changes (a new health check, say). Upgrade as a separate step, from the console, when you are ready for that.
• Keep the database volume. Without pvx-provelox-data, Provelox starts empty and knows nothing about the apps.
• Before you start: write down any app secret you may need again (Provelox cannot read them back), back up the data volumes you care about, and run each command without --yes first to read the plan.
• Try it on a copy first if you can. A throwaway server takes minutes to set up with Install.

Bring Provelox back

Run install again with the same root domain and the same image (add your other options, such as --acme-email, --http-port):

docker run --rm -v /var/run/docker.sock:/var/run/docker.sock \
  ghcr.io/provelox/provelox:latest install --root-domain example.com --acme-email you@example.com

When it finds pvx-provelox-data it keeps it:

Found the data of an earlier install and kept it: sign in with the admin password you already had. Your apps are
re-created from it, with their settings and variables. An app secret Docker no longer holds (after leaving the Swarm,
or `uninstall --remove-apps --delete-data`) has to be set again.

On the test server the old admin password signed in again, an app came back with its environment variable and its secret, its error-page step was restored, and an app's data volume was untouched. This only works if you kept pvx-provelox-data. Without the data volume Provelox starts empty and knows nothing about the apps, which keep running as plain services but are no longer managed.

By hand

On a version older than 0.5.0, or to see exactly what the commands do: list the objects (docker service ls, docker network ls, docker volume ls; everything Provelox made starts with pvx- or provelox-), remove the services (docker service rm pvx-console pvx-provelox pvx-traefik pvx-<app> …), then the secrets, configs and networks, and the volumes last. For the keep-the-apps case, first remove the error-page step from each app's routes: for every traefik.http.routers.*.middlewares label drop the entry named <app>-errors, and remove the traefik.http.middlewares.<app>-errors.* labels, with docker service update --label-rm and --label-add.