Verification

osd/ocp cluster

run against cluster:

oc new-project apicast-test

make install

make run

on another namespace install 3scale operator (with APIM etc, either from local or operators hub)
create a secret pointing to the 3scale admin portal to be consumed by the apicast CR on the same namespace where apicast operator is watching
```
kubectl create secret generic my-test --from-literal=AdminPortalURL=https://<ADMIN_TOKEN>@<ADMIN ROUTE - without https://>
```

create apicast CR:

apiVersion: apps.3scale.net/v1alpha1
kind: APIcast
metadata:
name: example-apicast
namespace: apicast-test
spec:
adminPortalCredentialsRef:
name: my-test
replicas: 4
resources:
limits:
  cpu: 2000m
  memory: 500Mi
requests:
  cpu: 300m
  memory: 150Mi

Wait for apicast CR to report completed
The outcome here is that apicast pods will start up, with 4 replicas and above limits and request
Set the hpa: true on apicast CR
Outcome will be that the hpa object will be created, with 1-5 pods and initial value of pods matching the previous value which was - 4
Let it run for about 10 minutes, the number of replicas should reduce over time
Confirm that a warning condition is visible on the apicast CR indicating that requests, limits, and replicas values are ignored because HPA took over.
Confirm that requests and limits are set to the same values

    cpuLimits := resource.MustParse("1")
    memoryLimits := resource.MustParse("128Mi")

Set the HPA to false / or remove it. The resource limits, requests and replicas should go back to old values (or the ones specified in the APICast CR) and the warning message should be removed from the CR.

MStokluska commented 10 months ago

TODO: docs to be updated post final review

austincunningham commented 10 months ago

Verification

[x] Apicast created with 4 replicas
[x] Deployment had the specified requests and limits

[x] Set spec.hpa: true and the deployment requests == limits and after about 10 min the replicas scale down

oc get hpa
NAME              REFERENCE                            TARGETS           MINPODS   MAXPODS   REPLICAS   AGE
example-apicast   Deployment/apicast-example-apicast   23%/90%, 0%/90%   1         5         1          10m

Weird issue in the ui seems to always have one pod pending

[x] confirm the message in the CR

status:
conditions:
- lastTransitionTime: '2024-01-18T14:28:53Z'
  message: APIcast is ready
  reason: Ready
  status: 'True'
  type: Ready
- lastTransitionTime: '2024-01-18T14:28:53Z'
  message: >-
    HorizontalPodAutoscaling (Hpa) enabled overrides values applied to
    request, limits and replicas
  reason: HPA
  status: 'True'
  type: Warning
image: 'quay.io/3scale/apicast:latest'
observedGeneration: 5

[x] remove spec.hpa: true and the pods scale up and deployment requests and limits match those specified in the CR

NOTE: pending pods go away, but will appear again once hpa in enabled, don't believe its related to this PR . Wonder if its an issue with the HPA controller

austincunningham commented 9 months ago

code review /lgtm /approve

3scale / apicast-operator

THREESCALE-10366 initial config of HPA for Apicast Operator [WIP] #207

Verification

Verification