Pipeline stucks for kubernetes if image is invalid

[zc-devs]

Component

server

Describe the bug

When the one forgot commands directive, then they join to image => image cannot be pulled, pipeline stucks.

Steps to reproduce

1. Create and run pipeline

skip_clone: true

when:
  - event:
      - manual

steps:
  hello:
    image: alpine
      - echo Hello
      - echo world

2. See that it runs without errors and warnings
   Screenshot 2024-06-29 1
   Screenshot 2024-06-29 2
3. Pipeline stucks until timed out or cancelled manually
4. Check Pods and see errors

Error: InvalidImageName

Screenshot 2024-06-29 3

Failed to apply default image tag "alpine - echo Hello - echo world": couldn't parse image name "alpine - echo Hello - echo world": invalid reference format: repository name (library/alpine - echo Hello - echo world) must be lowercase

Screenshot 2024-06-29 4

Expected behavior

Pipeline should fail and linter should warn about missed commands as well.

System Info

`2.6.0`, Kubernetes

Additional context

woodpecker-server.log
woodpecker-agent.log

Validations

• Read the docs.
• Check that there isn't already an issue that reports the same bug to avoid creating a duplicate.
• Checked that the bug isn't fixed in the next version already [https://woodpecker-ci.org/faq#which-version-of-woodpecker-should-i-use]

[zc-devs]

In regard Kubernets backend, probably, wider range of statuses should be hacked (#3580).

[qwerty287]

Not sure, but this seems to be a yaml issue. AFAIK docker backend tells you if it fails to pull the image, or do you only specifically about kubernetes?

[zc-devs]

yaml issue

It is. It is only a part. Second is the Server's checks/linter. And third is Kube backend.

docker backend tells you if it fails to pull the image

Cool :) Kubernetes stucks.

do you only specifically about kubernetes?

There, where backend is meant, yes, I'm about Kubernetes backend, cause I have no other backends.

---

#3909 could have fixed the Server part. Schema is right, but there are no checks for commands and entrypoint steps.

[6543]

well your yaml is parsed:

skip_clone: true

when:
  - event:
      - 'manual'

steps:
  hello:
    image: 'alpine - echo Hello - echo world'

so the error that the image is invalid is fine!

how should the linter know if it is not a plugin and misses commands ?!?

[zc-devs]

the error that the image is invalid is fine!

Absolutely, if pipeline fails (third part).

how should the linter know if it is not a plugin and misses commands ?!?

I thought, that plugin uses settings. You are right:

"required": ["image"]

However, we check volume/pod/service names in Kube backend. Some regexp could be introduced for image check in the server.

[6543]

well that's an enhancement proposal :)

#3919

[zc-devs]

What about

Pipeline stucks until timed out or cancelled manually
In regard Kubernets backend, probably, wider range of statuses should be hacked (#3580).

?

What if I am just user without access to the cluster?

[6543]

uh missed that, this is another issue!