You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
We tested capacitor on one of our k8s clusters and figured out, that capacitor uses a lot of memory. Our cluster has regarding to flux the following basic data:
sources: 1
kustomizations: 2
helm releases: 67
We are using capacitor v0.3.2 with flux v2.2.3 and kubernetes v1.27.5
Grafana gives us the following memory related graph:
The first and the last peak are single reload requests on the ui, the middle part are 3-times 3 parallel reload requests.
Normal behavior or bug?
The text was updated successfully, but these errors were encountered:
Thank you for this great product.
We tested capacitor on one of our k8s clusters and figured out, that capacitor uses a lot of memory. Our cluster has regarding to flux the following basic data:
We are using capacitor v0.3.2 with flux v2.2.3 and kubernetes v1.27.5
Grafana gives us the following memory related graph:
The first and the last peak are single reload requests on the ui, the middle part are 3-times 3 parallel reload requests.
Normal behavior or bug?
The text was updated successfully, but these errors were encountered: