Cluster Upgrade Stuck – Node Version Conflict Causing "No agent available" & Pod Failures

[S-GOKUL11]

Kubernetes cluster upgrade is stuck in an inconsistent version state.

Some nodes are on version 1.32.9 while the system is trying to enforce 1.31.1.
Admission webhook is blocking operations with error:

"newVersion 1.31.1 is not allowed for oldVersion 1.32.9"

Cluster is partially upgraded and workloads are failing.
Unable to reboot or replace nodes due to version downgrade restriction.

Requesting:

• Force completion of upgrade OR
• Version reconciliation across node pool OR
• Manual intervention to fix version mismatch

[hsinghal-pf9]

Hi @S-GOKUL11 , thanks for reaching out.
Could you please share your organization name and cloudspace name so that we can investigate this issue.

[S-GOKUL11]

Sure @hsinghal-pf9 ,

Organization Name : sbnatechnologies
Cloudspace Name: sbna

[sirishkr]

@S-GOKUL11 can you help us understand: did you attempt to "upgrade" to an older version here - 1.31.1?

[S-GOKUL11]

@sirishkr

Sorry for the confusion — we did not manually perform any upgrade/downgrade. The cloudspace was recreated in an emergency due to a cluster issue.

---

Current Cloudspace Details

• Org: sbnatechnologies
• Cloudspace: rcmus (recreated)
• Node type: Extra Large (8 vCPU, 30 GB RAM, 40 GB Disk)
• Kubernetes version: v1.33.0
• Nodes: 5 worker nodes (Spot instances)
• Status: All nodes are Ready

---

Observation

One of the worker nodes was provisioned with 1 vCPU instead of the expected 8 vCPU from the start.

---

Request

Could you please confirm:

• Why one node is provisioned with 1 vCPU instead of the configured 8 vCPU
• Whether this is expected in Spot instance provisioning or a node pool configuration issue
• If the node should be replaced/recreated to match the correct sizing

Thanks.

[S-GOKUL11]

Hi @sirishkr @hsinghal-pf9 @szilveszter @cloudnull,

Could you please take a look at this when you get a chance?

The cloudspace was recreated due to an emergency cluster issue, and we've observed that one of the worker nodes was provisioned with only 1 vCPU, while the selected node type is Extra Large (8 vCPU, 30 GB RAM).

We would appreciate clarification on:

• Why this node was provisioned with 1 vCPU instead of the expected 8 vCPU.
• Whether this behavior is expected when using Spot instances.
• If this indicates a node pool provisioning issue.
• Whether the node should be replaced/recreated to match the configured sizing.

This is currently impacting our capacity planning, so any guidance would be greatly appreciated.

Thank you.

[hsinghal-pf9]

cc @vjha-pf9

[vjha-pf9]

@S-GOKUL11 Thanks for reporting this issue. Ideally, this should not happen in a spot. The issue seems to be linked to Rackspace Infra, where due to some issue, the wrong node with one CPU was provisioned prod-instance-17751181276201068. We have recycled it and attached the new node to your cloudspace. Please check and confirm

[S-GOKUL11]

@vjha-pf9 Thank you for the update and for resolving the issue.

We have verified the new node and can confirm it is provisioned with the expected resources. Appreciate the quick investigation and support.