Strange behavior of ensemble placement policies

**QUESTION**

We use BookKeeper with Pulsar. We tried enabling the ensemble placement policy on Pulsar's broker and we saw some strange behavior. Please tell me if these behaviors are as expected.

1. If the policy is `RackawareEnsemblePlacementPolicy`, bookies with rack information and bookies without rack information cannot coexist in the same cluster. If they co-exist, we get the following error:
```
11:36:13.177 [BookKeeperClientScheduler-OrderedScheduler-0-0] WARN o.a.b.c.RackawareEnsemblePlacementPolicyImpl - Failed to resolve network location for dev-bookie101.pulsar.xxx.xxx.yahoo.co.jp, using default rack for it : /default-rack.
11:36:13.177 [BookKeeperClientScheduler-OrderedScheduler-0-0] ERROR o.a.b.net.NetworkTopologyImpl - Error: can't add leaf node at depth 2 to topology:
Number of racks: 2
Expected number of leaves:2
/region1/rack1/dev-bookie100.pulsar.xxx.xxx.yahoo.co.jp:3181
/region1/rack2/dev-bookie102.pulsar.xxx.xxx.yahoo.co.jp:3181
```

2. If the policy is `RackawareEnsemblePlacementPolicy`, adding or removing bookie rack information does not take effect dynamically (until the broker or the bookie is restarted). For example, even if we add rack information to a bookie that did not have it, that bookie will not be selected as an ensemble member.

3. If the policy is `RegionAwareEnsemblePlacementPolicy`, region information updates does not take effect dynamically (until the broker or the bookie is restarted).