[Autoscaling] Add setting to customize autoscaling API polling frequency #4264
Conversation
barkbay
added
>feature
| @@ -52,6 +57,10 @@ type AutoscalingPolicy struct { | |||
| // +kubebuilder:object:generate=false | |||
| type AutoscalingSpec struct { | |||
| AutoscalingPolicySpecs AutoscalingPolicySpecs `json:"policies"` | |||
|
|
|||
| // SyncPeriod is the period at which to synchronize and poll the Elasticsearch autoscaling API. | |||
| SyncPeriod *metav1.Duration `json:"syncPeriod"` | |||
There was a problem hiding this comment.
Should it simply be called pollingFrequency or pollingPeriod? If I read sync I wonder what it is being synchronized with what.
| var defaultRequeue = reconcile.Result{ | ||
| // licenseCheckRequeue is the default duration used to retry a licence check if the cluster is supposed to be managed by | ||
| // the autoscaling controller and if the licence is not valid. | ||
| var licenseCheckRequeue = reconcile.Result{ |
There was a problem hiding this comment.
What is the motivation for using a different interval on invalid licenses? To avoid a bad experience for users that set syncPeriod to a really high value? Initially I thought it would be to protect the operator from users that set it to a really low value because the license check is relatively expensive. But I realise that the check runs on every reconciliation loop anyway. Maybe worth a comment.
There was a problem hiding this comment.
To avoid a bad experience for users that set syncPeriod to a really high value?
Yes exactly, my initial thought was that if the license is invalid it might be a user mistake, and we may want to trigger a new loop sooner than later, regardless of the syncPeriod value which serves an other purpose.
barkbay
force-pushed
the
autoscaling-pr/syncperiod
branch
from
36ebde8
to
802fc67
Compare
This PR allows the user to customize the default autoscaling API polling frequency.