Awkward pod status message for admission errors due to node-pressure condition

[mimowo]

What happened?

When a Pod admission fails due to node-pressure condition then the Pod's status.message looks like this:
"Pod The node had condition: [DiskPressure]."
The "Pod The node..." reads awkward.

What did you expect to happen?

A message like: "The node had condition: [DiskPressure].".

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

1. Run a job with a pod on a single node cluster, when the node is under disk pressure (then it is tainted with node.kubernetes.io/disk-pressure:NoSchedule). The Pod gets stuck in the Pending phase.
2. Untaint the node by kubectl command line: kubectl taint node $nodeName node.kubernetes.io/disk-pressure:NoSchedule-
3. The Pod is scheduled but admission by Kubelet fails as the taint is re-added shortly after its manual removal.

Then, the status of the Pod contains the awkward message.

Anything else we need to know?

No response

Kubernetes version

$ kubectl version
1.25

Cloud provider

Reproducible on kind

OS version

# On Linux:
$ cat /etc/os-release
# paste output here
$ uname -a
# paste output here

# On Windows:
C:\> wmic os get Caption, Version, BuildNumber, OSArchitecture
# paste output here

Install tools

Container runtime (CRI) and version (if applicable)

Related plugins (CNI, CSI, ...) and versions (if applicable)

[SergeyKanzhelev]

/sig node
/triage accepted
/priority backlog

[vitorfhc]

The first part (Pod The node had condition: [DiskPressure]) is returned here: eviction_manager.go.

I wonder where the Pod is being appended to the beginning. It's my first try to contribute, so I am a little bit lost.

[mimowo]

The first part (Pod The node had condition: [DiskPressure]) is returned here: eviction_manager.go.

I wonder where the Pod is being appended to the beginning. It's my first try to contribute, so I am a little bit lost.

I think this is a candidate place which might be responsible for the prefix (but haven't checked):

kubernetes/pkg/kubelet/kubelet.go

Line 1963 in 17c5066

Message: "Pod " + message})

I would suggest you to try to reproduce the issue locally first then experiment with modifying the code.

[vitorfhc]

/assign

[vitorfhc]

@mimowo thanks, I've managed to reproduce and used Delve to make sure who's responsible.

I opened the PR changing the message. If it's accepted it will change to Pod rejected: The node had condition: [DiskPressure].