Value of ephemeral container property targetContainerName not present on host

[dee0sap]

What happened?

We have enabled ephemeral container syncing

sync:
  pods:
    enabled: true 
    ephemeralContainers: true

Within our vcluster we have set values for .spec.ephemeralContainers[].targetContainerName however when we examine the pods that syncing created on the host targetContainerName is unset.

See example yaml in attached zip.
pod-target-has-multiple-mounts0-redacted.zip

What did you expect to happen?

I expect the pod on the host to have the same targetContainerName value as the pod in the vcluster.

How can we reproduce it (as minimally and precisely as possible)?

• Enable syncing as above
• Use 'kubectl debug' with the '--target' option

Anything else we need to know?

Not sure if it is related, but will mention for completeness' sake.

kubectl debug creates ephemeral containers via PATCH. I have code that is doing this programmatically and I found that rapidly PATCHing in effort to add a pair of ephemeral containers causes the pods on the host and the vcluster to get out of sync and/or corrupted. Generally 'out of sync' would look like

• ephemeral container missing from host
• resource version data suggesting that syncer things the objects are in sync when they are not
  Corruption would look like
• Possibly an ephemeral container missing from the pod on the host
• On host ephemeralContainerStatuses show that the ephemeral containers all using same image when they should have been using different ones

Host cluster Kubernetes version

Client Version: v1.30.2
Kustomize Version: v5.0.4-0.20230601165947-6ce0bf390ce3
Server Version: v1.29.3

vcluster version

$ vcluster --version
vcluster version 0.19.6

VCluster Config

sync:
  priorityclasses:
    enabled: true
  pods:
    enabled: true 
    ephemeralContainers: true
syncer:
  extraArgs:
  - --tls-san=XXXX
init:    
  manifests: |-
   # A priority class definition 
    ---
   # A secret for docker image pull

Above is the config. Note that I redacted a few things.

[deniseschannon]

We've made some refactoring in the code around syncing in our v0.21.0 release version, but it's still in alpha.

Would you be willing to test out our latest alpha (v0.21.0-alpha6) if this is still an issue in v0.21.0?

[dee0sap]

Hey @deniseschannon

Sure. Btw, I might not be able to test it until next week though.

[dee0sap]

Hey @deniseschannon
Finally got to try this. Tried vcluster version 0.21.0-beta.2 and it is fixed.

[cbron]

Thanks for testing @dee0sap , if this is fixed can it be closed ?

[dee0]

Hey @cbron

Yep, this can be closed. Thanks!