In my experience with different Kubernetes services and BYO solutions there's disparity in between vendors and in-tree/out-of-tree drivers. It could be how effective PVC creation and subsequent PV binding perform. What also differs is how reliable attach/detach is during node drains or mass deployment, which sometimes fails altogether.
"Storage is hard"
... is the most used term in this space and it would therefor be very handy to use kboom as a sizing/discovery tool for what the optimal PVC/PV count is for a certain cluster and node size. Once you have an idea of where the boundaries are the cluster-admin can then divide volume counts across the projected namespaces and therefor make the cluster safer and won't run off a cliff blindly when not knowing where the ceiling is for the particular environment.
In my experience with different Kubernetes services and BYO solutions there's disparity in between vendors and in-tree/out-of-tree drivers. It could be how effective PVC creation and subsequent PV binding perform. What also differs is how reliable attach/detach is during node drains or mass deployment, which sometimes fails altogether.
... is the most used term in this space and it would therefor be very handy to use kboom as a sizing/discovery tool for what the optimal PVC/PV count is for a certain cluster and node size. Once you have an idea of where the boundaries are the cluster-admin can then divide volume counts across the projected namespaces and therefor make the cluster safer and won't run off a cliff blindly when not knowing where the ceiling is for the particular environment.