ロードマップ

`droonga-engine`

Make drop-in replacement of log search system based on Groonga.
Optimize performance of search for unsharded replicas. (Currently, mechanisms of hash-based sharding always affect for replicas. They should be skipped for unsharded replicas.)

Support hash-based sharding of replicas, by droonga-enigine-join and droonga-engine-unjoin commands.
Support non-stop modification of cluster structure. (Currently, we have to stop updating data while modifying cluster structure. Joining of new replica should be done completely silently. To do it, we have to do: 1) half-unjoin a source replica from the cluster, 2) synchronize data from the source replica to the new destination replica, 3) join those replicas to the cluster, and 3) wait to dispatch search requests to newly joined replicas until buffered updating requests are completely processed.)
Support various type system notifications. (Currently, munin or something can be available to monitor status of nodes. However it should be integrated.)
Better compatibility to Groonga: Support suggest and related commands.
Support various type shardings (e.g. time-range based).

Detect available droonga-engine nodes automatically. (Nodes should fetch catalog.json from initially connected droonga-engine node via the catalog.fetch command and refresh connections based on the cluster definition.
Prevent to send request messages to dead droonga-engine nodes. (Nodes should join to the orchestration based on Serf.)

Better dashboard. (For example, search columns should be listed and filtered based on the schema information. This better dashboard will be developed in a separate repository to be shared by Groonga and Droogna…)