If your deployment distributes entries using an entry counter placement algorithm or 3rd party algorithm, you may need to redistribute your entries. For example, imagine that you have an environment that distributes entries across three backends using an entry counter placement algorithm. This algorithm distributes entries to the backend that has the most space. Imagine that the backends all reach their maximum capacity and you decide to add a new backend to the deployment. You need to move the entries from the full backends and distribute them evenly across all the backends, including the new backend.
You might also want to deliberately rebalance your entries to meet the needs of your organization. For example, you can direct entry balancing based on attributes on the entries themselves. You can write a custom algorithm that looks at the value of an attribute that is being modified on the entry. Based on the attribute, you can then put this entry somewhere specific. You might use this feature if you want to have certain entries closer geographically to the client application using them. The geographical information could be included in the entry. Rebalancing would be used to move these entries to the server in the correct geographical location.
You can redistribute entry-balanced entries in two ways:
- Using dynamic rebalancing. With dynamic rebalancing, as existing entries get modified, they get moved. You configure dynamic rebalancing in the entry counter placement algorithm.
- Using the move-subtree tool. This tool can be used to move either small subtrees through a transactional method or to move large subtrees, potentially taking them offline for a short period.
The remainder of this section describes each of these method of entry rebalancing in more detail.