kcl v0.10.10 is failing to start

[fraenkel]

What happened?

Upgrading to 0.10.10

{"level":"info","ts":1733853420.7700005,"caller":"fn/fn.go:37","msg":"Running Function","tag":""}
{"level":"info","ts":1733853420.7716818,"caller":"fn/fn.go:37","msg":"Running Function","tag":""}
{"level":"info","ts":1733853420.7729983,"caller":"fn/fn.go:37","msg":"Running Function","tag":""}
2024/12/10 17:57:00 failed to write file:  /tmp/kcl.mod open /tmp/kcl.mod: permission denied

How can we reproduce it?

Just switched from 0.10.8 to 0.10.10

What environment did it happen in?

Function version: 0.10.10

I see there was a previous issue which hit a similar issue over the KCL_CACHE_PATH.
The current image has KCL_CACHE_PATH=/tmp

[fraenkel]

So the v0.10.10 images have files in /tmp whereas they didn't before.

root@2066a6833416:/# ls /tmp
kcl.mod  kcl.mod.lock

[irizzant]

Hitting the same here, v0.10.9 works fine

[vfarcic]

Hitting the same as well.

[dmvariphy]

We're seeing this as well with v0.10.10.

EDIT: We did see the v0.10.10 function go ready after deleting a Configuration package - not sure if that had anything to do with it, or if the issue is intermittent. The function had been restarting for almost an hour before that though.

EDIT2: If it helps at all, we see the issue when running kind on wsl, but not when running kind on linux.

[Peefy]

cc @zong-zhe @He1pa

[vfarcic]

Can you remove the latest release until this is resolved? Otherwise, it'll fail for anyone using >= as the dependency version which results in the latest release of the function being deployed.

[zong-zhe]

Can you remove the latest release until this is resolved? Otherwise, it'll fail for anyone using >= as the dependency version which results in the latest release of the function being deployed.

Sorry😞, I'm on my way home. This release will be removed soon when I get home.

[zong-zhe]

Hi @vfarcic, @fraenkel, @irizzant, @dmvariphy

I've rolled back version 0.10.10 to a pre-release to prevent this issue from blocking your work.

[dmvariphy]

Thanks for doing that @zong-zhe!

[dzmoore]

Hey @zong-zhe, I'm still pulling xpkg.upbound.io/crossplane-contrib/function-kcl:v0.10.10 with a >= dependency in my crossplane.yaml file (I think this is what @vfarcic was referring to in #206 (comment)). e.g.

apiVersion: meta.pkg.crossplane.io/v1
kind: Configuration
# ...
spec:
  # ...
  dependsOn:
   # ...
    - function: xpkg.upbound.io/crossplane-contrib/function-kcl
      version: ">=v0.9.0"

$ k logs -n crossplane-system -l pkg.crossplane.io/function=crossplane-contrib-function-kcl
{"level":"info","ts":1734388399.8846216,"caller":"fn/fn.go:37","msg":"Running Function","tag":""}
2024/12/16 22:33:19 failed to write file:  /tmp/kcl.mod open /tmp/kcl.mod: permission denied

$ k get functions.pkg.crossplane.io crossplane-contrib-function-kcl 
NAME                              INSTALLED   HEALTHY   PACKAGE                                                    AGE
crossplane-contrib-function-kcl   True        False     xpkg.upbound.io/crossplane-contrib/function-kcl:v0.10.10   6m21s

Is there any chance that the package is still in the registry (maybe needs to be removed), or is there maybe a cache I need to clear locally?

I can work around it for now by manually changing crossplane.yaml and re-creating my Kind cluster, but just wanted to mention we're running into this.

[vfarcic]

Using >= as a dependency version is common in Crossplane meaning that many users of KCL functions are potentially affected. Please remove the latest release from the marketplace as a hot-fix.

[irizzant]

I agree, my renovate bot still shows 0.10.10 as available

[zong-zhe]

Hi @irizzant, @vfarcic, @dzmoore, @fraenkel, @dmvariphy

The 0.10.10 will be removed soon.

[vfarcic]

It works. Thanks a ton @zong-zhe.

[mproffitt]

Pulling v0.10.10 has caused issues when using OCI images and I will also presume github due to missing CA certificates.

This is a regression of #191 which was fixed in 0.10.10 and will affect some users.

Side question - is it worth pinning this issue for visibility until a new release is out that replaces .10 or is a fix expected soon?

[zong-zhe]

Pulling v0.10.10 has caused issues when using OCI images and I will also presume github due to missing CA certificates.

This is a regression of #191 which was fixed in 0.10.10 and will affect some users.

Side question - is it worth pinning this issue for visibility until a new release is out that replaces .10 or is a fix expected soon?

Hi @mproffitt

We have found the issue and are working on fixing it for 0.10.10; soon, 0.10.10 will be re-released.

[zong-zhe]

Hi @fraenkel @irizzant @vfarcic @dmvariphy @dzmoore @mproffitt 😄

I republished 0.10.10. If there is any problem, @ me here, thanks !