Editing Deployment image names via "Edit Config" results in failure

[nickgerace]

When editing a deployment's image name, via "Edit Config" errors occur. However, when editing it via "Edit YAML", no errors occur.

Error that appears in UI:

Deployment.apps "fleet-agent" is invalid: spec.template.spec.affinity.nodeAffinity.requiredDuringSchedulingIgnoredDuringExecution.nodeSelectorTerms: Required value: must have at least one node selector term

Changing this one field appears to create a large diff as well when using "Edit Config". One field is offset too:

[nwmac]

@Sean-McQ Check with @rak-phillip - I think he looked at a similar issue elsewhere (I could be wrong)

[gaktive]

There are additional reports about this happening. Internal reference: SURE-3676.

[gaktive]

From the internal reference, there's a comment in place tied to preference that may be the culprit.

[ghost]

Was unable to reproduce and after a brief conversation with @mantis-toboggan-md it looks like this was previously resolved in 2.6.3 under #4319

[sgapanovich]

Reproduction

Rancher version: 2.6.3
Rancher cluster type: single-node docker install

Downstream cluster type: rke2 ec2 node driver
Downstream K8s version: 1.21.9

1. Workloads -> Deployments -> Create
2. Name: test
   Container name: test
   Image: nginx
   Node Scheduling: Run pods on node(s) matching scheduling rules
   -Priority: Preferred
   -Key: node-role.kubernetes.io/worker
   -Operator: is set
   

---

Test Environment:

Rancher version: v2.6-head 4df2214
Rancher cluster type: HA
Docker version: 20.10

Downstream cluster type: rke1 ec2 node driver
Downstream K8s version: 1.23.4

Testing:

Edit Config for:

• deployment with Node Scheduling -> Run pods on nodes(s) matching scheduling rules -> Preferred (affinity)
• deployment with Node Scheduling -> Run pods on nodes(s) matching scheduling rules -> Required (affinity)
• deployment with Node Scheduling -> Run pods on specific node(s) -> Select node (nodeName)

All deployments are updated successfully with pods recreated