Pagination issues

[tuminoid]

These list calls do not even reach the API server since controller-runtime returns cached values. In addition this MR does not implement pagination, it just limits the amount of objects requested. If there is more than 200 resources to consider with one of the remaining list calls, this might lead to unintended behavior.

This change was reverted for Metal3Machine host association, but the remaining cases are lingering bugs for large-scale setups as well.

Originally posted by @schrej in #2646 (comment)

PTAL @fmuyassarov, @Sunnatillo

[Sunnatillo]

I can take a look.

[tuminoid]

/assign @Sunnatillo

[schrej]

Before you attempt to properly implement pagination - controller-runtime's cache doesn't support that.

I see two options:

1. optimize which resources are fetched (more precise filtering, indexes) to save memory
2. disable the cache for some resources and implement proper pagination

I would strongly prefer option one. The cache already takes care of reducing api server load, but there is room for improvement when listing resources to save some runtime memory (e.g. only listing available BMHs during scheduling)

[Rozzii]

Just for the record so far these two PRs have been merged to help with pagination:
#2918
metal3-io/ip-address-manager#1166
/triage accepted

[Rozzii]

/kind bug