Currently, there is a hard limit of 8GB for the etcd backend [1], which has become a limitation for the scalability required for large-scale or dense clusters.

[1] https://github.com/openshift/cluster-etcd-operator/blob/a060517d1452b3756b230de1d1393e3573de4f45/pkg/etcdenvvar/etcd_env.go#L42

This spike is to revisit the assumptions for historical reasoning behind the 8GB limit and see if those remain applicable today, and explore paths to make the ETCD_QUOTA_BACKEND_BYTES configurable over the API.

Considerations:

• What "pause" length would users experience with larger backend sizes during the defrag & compaction processes