Hi Juntao, Thanks for your contribution, you may already know that we introduced a proxy module in RocketMQ 5.0, some logic in the client should be moved to or shared with the proxy module in our design.
The feature you mentioned is also what the proxy module needs. More generally, the `reachable broker detector` is a special queue selector, the same as the queue selector for FIFO messages. So I wondered if maybe we could introduce a queue selector pipeline to support more similar demands, each selector will narrow the available message queue before sending messages out. Obviously, the queue selector pipeline can be reused in the proxy module easily. The similar demands that the selector pipeline may meet: 1. Detect the unreachable servers. 2. Select a specific message queue by message type. 3. Traffic management for producers, and maybe consumers also. 4. More features that could rich the ability of the proxy. Looking forward to more discussions about this RIP. Regards, Xinyu Zhou On Mon, Mar 27, 2023 at 11:58 AM fuyou <fuyou...@gmail.com> wrote: > +1 > > 季俊涛 <3160102...@zju.edu.cn> 于2023年3月27日周一 11:42写道： > > > Hi, RocketMQ Community: > > > > > > > > > > The current client module has built-in latency related functions, but > > this function is passive - isolating brokers based on the sending > results, > > and is not aware of their availability. When multiple brokers encounter > > problems, they may be randomly selected for transmission. Therefore, this > > RIP hopes to provide a client's ability to actively detect broker > > reachability. Subdividing the available states of brokers and turning on > > this function when the network fluctuates greatly will make the process > of > > selecting brokers more intelligent. So I would like to start an email > > thread to discuss RIP-63 Reachable broker detector in client. > > > > > > I have written my proposal and you can click on the link below： > > > > > https://github.com/apache/rocketmq/wiki/RIP-63-Reachable-broker-detector-in-client > > > > If you have any questions or suggestions, please reply to this email or > > comment on the proposal. > > > > Thanks > > Juntao Ji > > > > -- > ============================================= > > fuyou001 > Best Regards >
