hulk
> Additionally, the 'type' is decoded from iter->value(), and I don't quite understand why it requires scanning metadata when the entire implementation seems to be just a combination of scan...
> e.g. for DELETERANGE , we don't need to SCAN, instead we can just call DeleteRange, right? Yes, this would work if we don't support the `COUNT` option.
> We usually have COUNT to limit the work that the server should do (avoid long-running execution). @torwig We could simply use `rocksdb::DeleteRange` if we don't need to support the...
@daison12006013 Generally looks good, thanks for your contribution. It would be great if we could add dedicated methods for those exported fields instead of turning them to a public member...
@Rafiot Thanks for your report. Redis did have a different reply behavior when count is 1.
Yes, we can fix it to align with Redis.
@Rafiot Good point. It'd be better to document explicitly instead of silently breaking old behaviors in some programming languages. And always return the array format makes sense whether the count...
@jemc @geonove Kvrocks cluster mode isn't the same as Redis, and it uses https://github.com/apache/kvrocks-controller to manage the cluster information. So we don't expect to support the failover on the Kvrocks...
@jemc You don't need to use `CLUSTERX` command if you're running in HA mode. And you could just use the redis-sentinel's failover command to switch the role.
@jemc Thanks for your clarification. I'm sorry for misunderstanding that you're saying the cluster failover initially.