I have two questions: 1. we hope the CP part can be as powerful as istio's CP. If we want to achieve that, we need to write it in Go, so that we can reuse most of the libraries instead of writing our own one. 2. It's hard to find people good at writing Lua. If we write CP in Lua, it means both DP & CP require Lua developer.
Ming Wen <[email protected]> 于2021年3月29日周一 上午11:58写道： > > hello, dev, > > The current dashboard mangagr API, after about half a year of development, > is not easy to use in the following issues: > 1. Apache APISIX plug-in schema synchronization is not easy, which requires > that the DP and CP of Apache APISIX must use matching versions to work > together; > 2. custom plugins need to synchronize schema first then can displayed on > dashbaord; > 3. Most of the features of mangagr API are duplicated with Admin API; > 4. The mangagr API is not stable, and the code quality and code coverage > are not good; > > At the Apache APISIX meetup last Saturday, yuansheng and Junjie Ma and I > also discussed about this topic, and we came up with a new solution: > 1. For duplicate features between mangagr API and Admin API, mangagr API is > *directly* *forwarded* to Admin API; > 2. The mangagr API only implements a small part of the logic that is not > easy to handle due to the Lua language; > 3. Of course, this will also bring a new issue, that is, the dashbaord > project will be deployed together with Apache APISIX to work. > > This is the new flow chart: > Admin -> dashboard -> mangagr API -> Admin API by APISIX -> etcd > > This change will delete most of the mangagr API code, but it is transparent > to users. > It will solve the 4 issues mentioned above, and will introduce APISIX as > trade off. > > what do you think? > > Thanks, > Ming Wen, Apache APISIX PMC Chair > Twitter: _WenMing
