>> The goal is to make it possible to easily write ZooKeeper clients in non-JVM languages.
The proof of concept is still using jute as serialization format, which makes write a client library harder. Use protobuf as serialization format might achieve this goal the marshal / unmarshal code can be generated automatically. >> I don't know enough about the gRPC client/server implementations to know if it will meet the needs of ZooKeeper. We have done a similar project at Twitter where we added Finagle (Twitter's RPC library) network connection in addition to NIO and Netty server cnx supported by ZooKeeper. The motivation was to utilize the advanced feature such as load balancing and traffic control built in Finagle to improve ZooKeeper's reliability and performance. Our implementation is similar to this gRPC POC as both only use the RPC layer of the library, but still use jute as serialization format (as opposed to protobuf for grpc and Thrift for Finagle.). So technically, yes, this could be done. Though our solution worked to certain extent but we decided to continue use NIO/Netty for production and build load balancing / traffic control on top of it and take patches from FB/upstream. >> Then the question becomes what else do we fix at the same time? It's a good point, if we are free of backward compatibility concerns, then there are quite a few things to experiment. One idea is to get rid of jute, the other is to have a new way of session management - for example decouple session state with connection state. On Mon, Nov 18, 2019 at 4:40 PM tison <wander4...@gmail.com> wrote: > Interesting, at a first glance I thought it is a client wrapper for > polyglot communicating and > has nothing to do with ZOOKEEPER-102. Will take a look at the JIRA since > you attach > a plan of the whole story. > > Best, > tison. > > > Jordan Zimmerman <jor...@jordanzimmerman.com> 于2019年11月19日周二 上午5:23写道： > > > We can resurrect https://issues.apache.org/jira/browse/ZOOKEEPER-102 > > > > > On Nov 18, 2019, at 4:15 PM, Jonathan Wong <chi.jonathanw...@gmail.com > > > > wrote: > > > > > > I’d be willing to offer some of my spare time towards this effort. > > > > > > Jonathan Wong > > > > > >> On Nov 18, 2019, at 1:04 PM, Jordan Zimmerman < > > jor...@jordanzimmerman.com> wrote: > > >> > > >> A fresh look at APIs is definitely in order. So, this would be a > > potential ZK 4.x. > > >> > > >>> That said, we added things like rest in the past for similar reasons > > and it > > >>> never really took off... Would be a shame to see the same here. > > >> > > >> ZK has had a lot more activity with recent committers. So, maybe this > > time will be different? If at least one other person will sign up to work > > on it, I'll shepherd it. > > >> > > >> -JZ > > >> > > >>> On Nov 18, 2019, at 12:22 PM, Patrick Hunt <ph...@apache.org> wrote: > > >>> > > >>> There are quite a few benefits to using grpc imo. It's come up a few > > times > > >>> where I've been part of the discussion - ala we make it b/w compat it > > would > > >>> be a good move imo. Then the question becomes what else do we fix at > > the > > >>> same time? e.g. make version fields 64 bit rather than 32? etc... > > there are > > >>> a bunch of zk4 such jiras that could be addressed at the same time > (and > > >>> likely in a b/w compat way - ie zk3) > > >>> > > >>> That said, we added things like rest in the past for similar reasons > > and it > > >>> never really took off... Would be a shame to see the same here. > > >>> > > >>> Patrick > > >>> > > >>> On Mon, Nov 18, 2019 at 6:48 AM Jordan Zimmerman < > > jor...@jordanzimmerman.com> > > >>> wrote: > > >>> > > >>>>> That looks like great work. In order to address the issues, why not > > >>>> build on top of curator (https://curator.apache.org)? > > >>>> > > >>>> (Note: I'm the main author of Curator). I'd definitely try to make > > >>>> something like Curator for gRPC. I'm not sure exactly what that > means > > at > > >>>> this point. But, my main goal is to enable non-JVM clients. We have > C, > > >>>> python and few others now but they always lag with changes and are > > hard to > > >>>> maintain. > > >>>> > > >>>> -JZ > > >>>> > > >>>>>> On Nov 18, 2019, at 9:45 AM, Jörn Franke <jornfra...@gmail.com> > > wrote: > > >>>>>> > > >>>>>> That looks like great work. In order to address the issues, why > not > > >>>> build on top of curator (https://curator.apache.org)? > > >>>>> > > >>>>> I could support in case question rise with SASL, but I am not sure > > yet > > >>>> if I find the time to actively develop for this unfortunately > > >>>>> > > >>>>>> Am 18.11.2019 um 15:25 schrieb Jordan Zimmerman < > > >>>> jor...@jordanzimmerman.com>: > > >>>>>> > > >>>>>> Hi Folks, > > >>>>>> > > >>>>>> I've written a proof of concept implementation of a > > ServerCnxnFactory > > >>>> that implements gRPC. The goal is to make it possible to easily > write > > >>>> ZooKeeper clients in non-JVM languages. Using the proof of concept I > > was > > >>>> able to write a Golang client easily. What's the interest level of > > >>>> something like this? Let's discuss if it's worth pursuing. I'd be > > willing > > >>>> to move this from proof of concept to production but I'll need help > > (1 or 2 > > >>>> co-developers). > > >>>>>> > > >>>>>> If you want to try it, I've pushed the Golang client and some > > >>>> instructions here (let me know if you have any issues - I'm a go > > neophyte). > > >>>> Note: "zookeeper/test.go" is the interesting file: > > >>>>>> > > >>>>>> https://github.com/Randgalt/zkgrpc < > > >>>> https://github.com/Randgalt/zkgrpc> > > >>>>>> > > >>>>>> Here's the proof of concept on the ZK server side (the interesting > > >>>> files are RpcServerCnxn.java, RpcServerCnxnFactory.java, > > >>>> RpcZooKeeperServer.java and zookeeper.proto): > > >>>>>> > > >>>>>> > > >>>> > > https://github.com/apache/zookeeper/compare/master...Randgalt:wip-grpc < > > >>>> > > https://github.com/apache/zookeeper/compare/master...Randgalt:wip-grpc> > > >>>> > > >>>>>> > > >>>>>> Issues: > > >>>>>> Writing a client, even with gRPC, will require some work. Sessions > > have > > >>>> to be maintained, watchers have to be maintained, etc. > > >>>>>> Currently, Jute is deeply embedded in ZooKeeper. The proof of > > concept > > >>>> has to emulate Jute byte buffers. Ideally, this will be abstracted > so > > that > > >>>> only records could be used so that the gRPC connection doesn't have > > to keep > > >>>> marshalling/unmarshalling byte buffers > > >>>>>> I don't know enough about the gRPC client/server implementations > to > > >>>> know if it will meet the needs of ZooKeeper. Anyone have experience > > here? > > >>>>>> I haven't completely thought through how much work it will take to > > >>>> write useful clients. As I've shown with the proof of concept simple > > ZK > > >>>> CRUD db operations work well. I need to spend time writing a recipe > > such as > > >>>> Leader Election to see how much work is required. > > >>>>>> I'm not sure how things like SASL and reconfig would work with > gRPC > > >>>>>> > > >>>>>> -Jordan > > >>>> > > >>>> > > >> > > > > >
