Error running commands when using platform=podman

[jeffgus]

/kind bug

What versions of software are you using?

Operating System:
Fedora 40

Output of odo version:
odo v3.15.0 (10b5e8a)
also tried:
odo v3.16.1 (cd346c5-nightly)

How did you run odo exactly?

odo dev --platform=podman

Actual behavior

The pods startup in podman, but commands fail. I also get this error during startup:

✗ Building your application in container (command: install) [436ms]
Error occurred on Push - unable to log error unable to exec command [/bin/sh -c cd ${PROJECT_SOURCE} && (npm install) 1>>/proc/1/fd/1 2>>/proc/1/fd/2]: exit status 254: a running component clientapp doesn't exist on the cluster: json: cannot unmarshal array into Go value of type podman.PodInspectData

Expected behavior

odo startup should start podman pods without errors. Issuing odo run <command> should also work.

Any logs, error output, etc?

✗ Building your application in container (command: install) [436ms]
Error occurred on Push - unable to log error unable to exec command [/bin/sh -c cd ${PROJECT_SOURCE} && (npm install) 1>>/proc/1/fd/1 2>>/proc/1/fd/2]: exit status 254: a running component clientapp doesn't exist on the cluster: json: cannot unmarshal array into Go value of type podman.PodInspectData

[jeffgus]

I re-cloned the tree into a new dir. The start now works, but odo run is still failing.

Some files must have biffed up the initial sync.

[rm3l]

Thanks for reporting this issue, @jeffgus ! Would you mind sharing the output of odo version (to check the version of Podman)?
Also, the output of odo dev --platform=podman -v=9 please.
I suspect odo may no longer work with the newest versions of Podman..

[jeffgus]

I think you're right. Fedora 40 is using podman 5.1.1.

I also noticed that odo logs --follow --platform podman is not working. It doesn't kick out an error, it just sits there without any output.

odo v3.16.1 (cd346c5e6-nightly)

Server: https://xxxxxxxxxxxxxxxx:6443
OpenShift: 4.14.27
Kubernetes: v1.27.13+048520e
Podman Client: 5.1.1

I also tried with odo 3.15.0

The -v=9 produced A LOT of output. I'll make a simple devfile.yaml and get the output instead of the one I'm using which is more complicated.

[github-actions]

A friendly reminder that this issue had no activity for 90 days. Stale issues will be closed after an additional 30 days of inactivity.

[too-common-name]

Hi guys,
I have the same problem I think. My environment:

OS: Fedora 40
odo v3.16.1 (v3.16.1)
Podman Client: 5.2.3

I am able to run: odo dev --platform podman and all works fine. Once the pod is up and running if I execute:
odo run debug --platform podman
odo logs --platform podman
I got errors.
I think the problem is related to this PR: from several months the command podman pod inspect ... returns a json array instead a json object. I handle this case in a fork of odo, you can find it here. What do you think?

[rm3l]

Thanks @too-common-name for looking into this. Indeed, odo relies on the output of podman pod inspect, and it seems its output was changed in Podman 5+.
Would you mind contributing your changes back as a pull request here?
Thanks.

[rm3l]

/lifecycle active

[too-common-name]

Hi @rm3l,
I checked the pipeline results and fixed the variable shadowing issue. However, I’m not sure which integration tests are failing. When I run the tests locally, I get the following error:
[odo] ✗ unable to access podman. Do you have podman client installed and configured correctly? cause: exit status 125
Despite this, odo version correctly shows the Podman version, and I can run Podman commands without issues. Any guidance on troubleshooting this would be appreciated.

[rm3l]

Hi @rm3l, I checked the pipeline results and fixed the variable shadowing issue. However, I’m not sure which integration tests are failing. When I run the tests locally, I get the following error: [odo] ✗ unable to access podman. Do you have podman client installed and configured correctly? cause: exit status 125 Despite this, odo version correctly shows the Podman version, and I can run Podman commands without issues.

What's the Podman version displayed by odo version?

Any guidance on troubleshooting this would be appreciated.

I think this is because of a breaking change in Podman. IIRC, the current integration tests on Podman rely on a namespacing feature in Podman, but which got dropped in Podman v4.5.0: containers/podman@c4fe0af (#17586)
I guess you have a recent version of Podman. If so, you can set the PODMAN_USE_NAMESPACES env var to false prior to starting the tests..

[too-common-name]

Hi @rm3l, I confirm the problem is the one you spotted. I looked a bit inside the code and found the variable: setting PODMAN_USE_NAMESPACES to false tests run. I also incremented a bit the maximum number of open file descriptors as I had errors related to this.