Description

We find the MongoDB controller often issues deletion even when the resources do not exist. For example, in `deleteCfgIfNeeded`:

The controller directly issues deletion of config sts without checking its existence (and ignores `IsNotFound`). When sharding is always disabled, such deletion makes no change to the cluster since the config sts never exists. Given that `Get` (from local cache) is usually much faster than `Delete` (to the apiserver), checking the sts existence by issuing `Get` before issuing `Delete` of the sts would be better. Note that in the same `deleteCfgIfNeeded`, svc is deleted in this manner:

In addition, this approach also helps to check the local indexer staleness. If `Get` confirms the resource existence but `Delete` returns `NotFound` error, it indicates potential staleness in the local indexer.

Similar problems also exist in other methods like `deleteMongos`.

Fix

We are willing to send a PR for this issue. As mentioned before, `Get` before `Delete` would be better here.