Just tried and it works at my side on Ubuntu. What I am doing:

1. Open a terminal, and start podman api service:

systemctl start --user podman.socket

2. Start onedev using same command as yours
3. Test server docker executor

thanks for your help.

I've just test in a fresh ubuntu server install

just install podman

# apt-get install podman

run socket as user

$ systemctl --user start podman.socket

run onedev

$ podman run -it --name onedev --rm -v $XDG_RUNTIME_DIR/podman/podman.sock:/var/run/docker.sock -v onedev_data:/opt/onedev -p 6610:6610 -p 6611:6611 docker.io/1dev/server

And another error, maybe i'm doing it wrong ??

hum, back on nixos just after the executor test, the podman.socket failed

i'm digging to find where the issue come from...

I am testing on Ubuntu 20.04. What is your ubuntu version?

i took the last one, 22.04 with podman 3.4.4.

With nixos, podman is 4.2.0

Not sure what's wrong at your side. Just started a AWS ubuntu 22.04, follow same steps as yours, and everything works fine.

You may also mount the podman socket as /var/run/docker.sock into a container with docker tool (docker.io/docker for instance), and then inside that container run docker ps to see if it works.

Same distro, same commands two results, that's crazy.

can you confirm :

• you use the server docker executor in the onedev jobs executor menu without any specific option, just like if it was docker not podman ?
• just install podman package and not other podman-* ?
• run podman as normal user ?

i will try new ubuntu install and see what's happen.

Remark : i run ubuntu inside virtualbox but it should not have any related problem

Could you explain how onedev (docker edition) is doing job executor ?

This is all I have done after lauching a ubuntu 22.04 from AWS:

$ sudo apt update
$ sudo apt install -y podman
$ systemctl --user start podman.socket
$ mkdir onedev
$ podman run -it --name onedev --rm -v $XDG_RUNTIME_DIR/podman/podman.sock:/var/run/docker.sock -v $(pwd)/onedev:/opt/onedev -p 6610:6610 -p 6611:6611 docker.io/1dev/server

Then setup OneDev, create a server docker executor, give it a name, with other settings untouched, and test it to be working fine.

on ubuntu 22.04 on virtualbox (windows underlying host). copy/paste your commands, setup onedev, create docker executor and test it

always error

i will try on another machine, bare metal...

i've test the podman.sock like you've proposed above.

The podman.sock is working as docker ps inside docker look good

Please exec into OneDev container and run some container to see if it works.

it works

Then can you please run the same failed command as printed in build log to see what happens?

so the output in onedev docker container (ubuntu server 22.04/virtualbox/windows)

Remark: the first command did not return nothing (the filter volume not match any result). without, the onedev container is found

The volume filter is a bug of podman:

https://github.com/containers/podman/issues/16019

Maybe here is another bug of podman failing to run container with volume and working directory specified. It will be great if you do some more tests to see what is the difference of your environment and AWS launched environment, and submit a bug report for podman.

The hello-world image must not be used for testing purpose as before !!

When using another, like node, everything works as expected on ubuntu/podman.

There's still an bug with the podman.socket in nixos. I've report the issue, waiting for help..

so, a workaround exists for nixos 22.05. The next nixos 22.11 is already patched.

Just one last issue : the echo hello world is working but seems to be stuck afterward

whereas it should go until busybox ?

Same problem when adding a build job in a test projet. The build is running never ending.

Do you think there is also another bug with podman ?

Seems like another bug of podman. You may check if the container is still running.

last update : all bugs fixed in the nixos unstable channel that will be the next stable v22.11.

thanks.