we are removing and adding thousands of documents on daily basis. It often happens that inserted documents have the same vector value like already existing records (could it be a problem that we insert identical embedding several times into HNSW?). Then we remove old set of documents entirely (but we can see that they are not removed immediately, we can search for them).
Additionanly we can observe that from 3 replicas we are using all of them reports different storage used: 77, 81, 86%. Is there a way we can try to “sync” it somehow? Regenerating all those data might be quite costly. Could you provide us a hint how to fix replicas’ state?