Recreate stopped containers

[pbaylas]

I have some containers that are stopped by default and only started when accessed (via swag ondemand). This works very nicely, with one exception: when I update my containers via dockcheck (using -s) it pulls the new image for the stopped containers, but does not recreate the container. When the container gets started by swag ondemand, it seems to still be using the old image.

For my purposes it would be fine if dockcheck just started the containers, because ondemand will just shut them down again. However, based on #240, it sounds like it might be better for others if dockcheck remembered the state of the container and stopped it after recreating.

[mag37]

Thank you for raising the issue!

I'm not familiar with ondemand but that sounds interesting. Your idea of keeping track of state sounds sensible.

What state is the container in when it's not active?
And how is that handled behind the scenes, so I can test a similar approach? Like docker stop/pause or docker compose stop/pause/down?
Inspecting the plugin it looks like the stop is this call:

self.docker_client.containers.get(container_name).stop()

There's a few things to inspect, if you just inspect the container docker inspect XYZ and look at the top section State or you could get the individual values with

$ docker inspect --format='{{ .State.Status }}' XYZ
running

$ docker inspect --format='{{ .State.Running }}' XYZ
true

There's currently a bit of work going on in #262 that's prioritized but will take a while because I'm personally choked (moving house + new job) but I'll get on testing with this when I find time! Sounds like a reasonable check to implement.

Edit:
The .State.Running is currently what's being tracked, but if there's a more granular info needed that can be obtained through any of the other .State-values (Status, Running, Paused) we'll use that too.

[pbaylas]

Thanks for the quick response! The containers are exited after they are stopped.

$ docker inspect --format='{{ .State.Status }}' dozzle
exited
$  docker inspect --format='{{ .State.Running }}' dozzle
false

[mag37]

Did a quick check now and if a container is stopped it's exited while if it's down we it dosn't exist so that'd throw an error.

~/serve/dozzle ➜ docker inspect dozzle -f '{{json .State }}'
{"Status":"running","Running":true,"Paused":false,"Restarting":false,"OOMKilled":false,"Dead":false,"Pid":1457027,"ExitCode":0,"Error":"","StartedAt":"2026-02-21T08:16:22.165305758Z","FinishedAt":"0001-01-01T00:00:00Z"}
~/serve/dozzle ➜ docker compose stop
[+] stop 1/1
 ✔ Container dozzle Stopped                                                                                                                                                                                                                                                                                             0.1s
~/serve/dozzle ➜ docker inspect dozzle -f '{{json .State }}'
{"Status":"exited","Running":false,"Paused":false,"Restarting":false,"OOMKilled":false,"Dead":false,"Pid":0,"ExitCode":0,"Error":"","StartedAt":"2026-02-21T08:16:22.165305758Z","FinishedAt":"2026-02-21T09:51:10.882114234Z"}
~/serve/dozzle ➜ docker compose up -d
~/serve/dozzle ➜ docker compose down
[+] down 2/2
 ✔ Container dozzle       Removed                                                                                                                                                                                                                                                                                       0.1s
 ✔ Network dozzle_default Removed                                                                                                                                                                                                                                                                                       0.0s
~/serve/dozzle ➜ docker inspect dozzle -f '{{json .State }}'
error: no such object: dozzle

Just noting this here if anyone else is experimenting (and for future me to work on this).

So if we'll track state - a error: no such object: xyz should mean downed/removed (no recreation) while exited should be consider a stopped but existing container.

[pbaylas]

I've had a go at implementing it in #269, feedback very welcome!

[mag37]

Nice one! Seems sensible, glanced through the code. I'll run it through some tests and consider if anything needs to be changed/rephrased - will give feedback if so. Thank you!