Panic on connection to podman dameon using TCP about pods

[K4rakara]

Upon inspecting the code, this appears to be happening because the only part of the port mapping that gets read is the protocol, not the host/container port, ip address, etc. I would say that would be the root issue, but on a local unix socket connection I don't have the same issue.

For example, on a container that was created with --publish 5173:80:

Some(
    PortMapping {
        container_port: None,
        host_ip: None,
        host_port: None,
        protocol: Some(
            "tcp",
        ),
        range: None,
    },
)

[marhkb]

Thanks for reporting.
Reading the documentation, it seems that Podman uses a random port if the field host_port is being omitted.

HostPort is the port number that will be forwarded from the host into the container. If omitted, a random port on the host (guaranteed to be over 1024) will be assigned.

But in this case, it shouldn't do this because of --publish 5173:80.

but on a local unix socket connection I don't have the same issue.

This is indeed weird. What version of Podman are you using?

[K4rakara]

I'm using podman version 4.3.1 on my laptop, 3.0.1 on my server. Perhaps thats the issue?

[marhkb]

Yeah, probably. The version is so old it even isn't listed anymore on the API reference overview :-)
https://docs.podman.io/en/latest/Reference.html

For my understanding: You are using the unix socket for 4.3.1 and the TCP socket for 3.0.1?

I think there wouldn't be an error if you'd expose the 4.3.1 via TCP podman instance and connect to that?

I will try to setup a VM with that particular podman version for reproduction purposes in the next days.

[K4rakara]

Correct, that's the setup (a certified Debian moment if I say so myself)

In the meantime I'll try to see if I can use a backport of a newer podman on the server.

[marhkb]

What you could also try is to change the problematic line

pods/src/model/container.rs

Line 331 in b036fd2

mapping.host_port.unwrap(),

To somethting like this

mapping.host_port.unwrap_or(80),

In this we, we can check whether more errors may follow

[K4rakara]

It looks like everything works after that! I can view the images, containers, processes, etc all just fine.

[marhkb]

Ok, I've set up the debian VM. The problem is that podman 3.0 uses camelCase for the "Ports" field names

❯ curl http://bullseye:37017/v3.0.1/libpod/containers/json
[{"AutoRemove":false,"Command":["/bin/sh"],"Created":"2023-01-30T20:58:12.329067206+01:00","CreatedAt":"","Exited":false,"ExitedAt":-62135596800,"ExitCode":0,"Id":"9c6a10c729bfbbc5d9fee00e645daa4839bfca68a5a425d8cd7fdf3e9fd8c517","Image":"docker.io/library/alpine:latest","ImageID":"042a816809aac8d0f7d7cacac7965782ee2ecac3f21bcf9f24b1de1a7387b769","IsInfra":false,"Labels":null,"Mounts":[],"Names":["flagrant-record"],"Namespaces":{},"Networks":["podman"],"Pid":797,"Pod":"","PodName":"","Ports":[{"hostPort":5555,"containerPort":5555,"protocol":"tcp","hostIP":""}],"Size":null,"StartedAt":1675108693,"State":"running","Status":""}]

Since 3.1 podman uses snake_case

"Ports":[{"host_ip":"","container_port":35017,"host_port":35017,"range":1,"protocol":"tcp"}],

See the documentation

Unfortunately, the library podman-api-rs, which Pod uses, doesn't cover this.

I can work around this specific issue by showing a "?" instead of an actual port number. But IMHO, it would be the best approach to upgrade to 3.1 since 3.0 is basically undocumented at this point.

[marhkb]

I decided to not show the port if it is absence. Please feel free to report all the bugs you encounter! Let's see how many of them we can actually fix.