MachinePool.Spec.ProviderIDList may not scale well
#6,387 opened on Apr 6, 2022
Repository metrics
- Stars
- (4,267 stars)
- PR merge metrics
- (PR metrics pending)
Description
What steps did you take and what happened:
In working on the MachinePool Machines proposal in #6088, we collectively rejected the idea of attaching an infrastructureRefsList to each MachinePool spec.
While this would be convenient, the overall size of a MachinePool could become large if the number of replicas / instances grows. Its representation is transferred over the network and stored in etcd. It's easy to imagine a long list of structs causing performance problems with a thousand-node cluster, for example. (No, we didn't test this, we are simply anticipating a possible problem.)
Similarly, the existing ProviderIDList on a MachinePool spec could become large in a large cluster. See this comment.
We should investigate if a large ProviderIDList actually causes problems–maybe this could be tested with DockerMachinePool? It might also be possible to deprecate use of ProviderIDList after #6088 has been implemented.
What did you expect to happen:
Anything else you would like to add:
Environment:
- Cluster-api version:
- Minikube/KIND version:
- Kubernetes version: (use
kubectl version): - OS (e.g. from
/etc/os-release):
/kind bug