Implement nodeSelector + taint support #40
Conversation
| } | ||
|
|
||
| // RunnerConfig defines the configuration of runner image | ||
| type RunnerConfig struct { |
There was a problem hiding this comment.
I think it a long time, that this should be the way to go; moving all configuration for runners in a separate struct. But I think it should be introduced as a new api version for this Kind (e.g v1alpha2) and provide a convertor from v1alpha1 to v1alpha2; otherwise it breaks backwards compatibility and existing SleepCycles from previous versions. If you try to run it on a cluster where existing SleepCycles exist, it will break:
A small example here: https://suedbroecker.net/2022/03/24/add-a-new-api-version-to-an-existing-operator/ (although the way it describes the changes that need to take place on the controller are a bit debatable and need more thorough analysis)
P.S: the project is not yet upgraded to kubebuilder v4 (still on v3; issue #29 which I am actively working on), so if the latest version of kubebuilder is installed on the development environment any attempt to add a new api will fail.
| // +kubebuilder:validation:ExclusiveMaximum=false | ||
| FailedJobsHistoryLimit int32 `json:"failedJobsHistoryLimit,omitempty"` | ||
|
|
||
| Runner RunnerConfig `json:"runner,omitempty"` |
| return &v1.ResourceRequirements{ | ||
| Limits: v1.ResourceList{ | ||
| "cpu": resource.MustParse("10m"), | ||
| "memory": resource.MustParse("128mb"), |
There was a problem hiding this comment.
That is not parsing, lines 132 and 136 need to change to 128Mi
2 participants
Implements #34
ResourceRequirementsTolerationsNodeSelectorPodMetadataPriorityClassNamerunAsUser