Thanks a lot and I think these could be very helpful ! Sheng Wu <wu.sheng.841...@gmail.com> 于2019年9月5日周四 下午10:30写道：
> We have a service comb RPC plugin, provided by you before. > For pack and transaction, we need another plugin. > > There are a lot of plugins you could ref, and Seata plugin is WIP too, > https://github.com/apache/skywalking/pulls. Wait for new protocol stable > at > Seata side. > > Sheng Wu 吴晟 > > Apache SkyWalking, Apache ShardingSphere(Incubating), Zipkin > Twitter, wusheng1108 > > > Zheng Feng <zh.f...@gmail.com> 于2019年9月5日周四 上午7:28写道： > > > So we have to dev a new plugin for the servicecomb-pack or there is a > > plugin we could leverage with ? > > > > Sheng Wu <wu.sheng.841...@gmail.com> 于2019年9月5日周四 下午10:08写道： > > > > > Then SkyWalking has a toolkit lib[1], you could use > > > `TraceContext.traceId()` in any place, if there is sw agent attached, > and > > > current context has been traced, you could have the trace id. > > > > > > [1] > > > > > > > > > https://github.com/apache/skywalking/blob/master/docs/en/setup/service-agent/java-agent/Application-toolkit-trace.md > > > > > > Sheng Wu 吴晟 > > > > > > Apache SkyWalking, Apache ShardingSphere(Incubating), Zipkin > > > Twitter, wusheng1108 > > > > > > > > > Zheng Feng <zh.f...@gmail.com> 于2019年9月5日周四 上午5:46写道： > > > > > > > Sheng Wu <wu.sheng.841...@gmail.com> 于2019年9月4日周三 上午11:31写道： > > > > > > > > > Willem Jiang <willem.ji...@gmail.com> 于2019年9月3日周二 下午6:25写道： > > > > > > > > > > > On Tue, Sep 3, 2019 at 10:52 PM Zhang Lei <zhang_...@boco.com.cn > > > > > > wrote: > > > > > > > > > > > > > > Hi Willem, > > > > > > > > > > > > > > I agree to integrate a transaction tracking service, I used > > > Zipkin, > > > > > but > > > > > > I don't know much about Skywalking. > > > > > > > > > > > > > > I have some questions: > > > > > > > > > > > > > > 1. Skywalking agent just working on the Omega side? > > > > > > Yep, but we can also let the agent do the instrumentation on the > > > Alpha > > > > > > side. > > > > > > > > > > > > > > > > > 2. Use @Trace in the same location as @SagaStart and > > @Compensable, > > > Or > > > > > > Combined annotation? And use TraceContext.traceId() to generate > > > > > globalTxId > > > > > > or localTxId ? > > > > > > This is the detail thing we need to think about. > > > > > > I think you are just proposing a way to integrate the Zipkin, am > I > > > > right? > > > > > > > > > > > > > > > > @Trace is SkyWalking annotation, which SkyWalking agent could > > identify > > > > and > > > > > instrument as a local span(Not RPC related) > > > > > > > > > > > > > > > > > > > > > > > 3. Will Omega users have conflicts if they already use Zipkin? > > > > > > We need to do some end to end verification, currently I think > it's > > > > > > fine if we can find a way to generate a correlation id between > > > tracing > > > > > > system and pack, we don't need to deeply combine these two system > > > > > > together, they can work separately. > > > > > > > > > > > > > > > > SkyWalking doesn't conflict with Zipkin, but I don't recommend to > use > > > > both > > > > > at the same time, because it is pointless. > > > > > > > > > > > > > > > > > > > > > > > 4. About provides a service to bridge the Trace information > that > > > > > > Skywalking collects with ServiceComb Pack Transaction events > > > > information. > > > > > > Do you mean that Alpha sends service chain information directly > to > > > > > > Skywalking via API? > > > > > > It's more like a portal integration. We can still leverage the > old > > > API > > > > > > provide by Skywalking and ServiceComb Pack. > > > > > > > > > > > > > > > > In this way, we need to have a deeper discussion about what > > ServiceComb > > > > > wants and requires. > > > > > > > > > Basically, I think we need to track the saga transaction status by > the > > > > global id. > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > Lei Zhang > > > > > > > > > > > > > > > 在 2019年9月3日，下午7:22，Willem Jiang <willem.ji...@gmail.com> 写道： > > > > > > > > > > > > > > > > Hi, > > > > > > > > > > > > > > > > As ServiceComb Pack use the GlobalTransactionId to trace the > > > > > > > > distributed transactions, it's make sense that let the > > > ServiceComb > > > > > > > > Pack and Skywalking share the same Id for tracing. > > > > > > > > In this way, we can bring the APM feature into Distribution > > > > > > > > Transaction managemet world. > > > > > > > > > > > > > > > > Current ServiceComb has Interface IdGenerator[1], which is > used > > > for > > > > > > > > the generate the GlobalTransactionId in OmegaContext[2]. I > > think > > > if > > > > > we > > > > > > > > can reuse the transaction implementation with the Skywalking > > > agent > > > > to > > > > > > > > generate the GlobalTransactionId. In this way we can search > the > > > > trace > > > > > > > > information and transactions with same correlation ID. > > > > > > > > > > > > > > > > The missing part is LocalTransactionId, we need to figure a > way > > > to > > > > > > > > bring the LocalTransactionId with the trace information. > > > > > > > > > > > > > > > > BTW, ServiceComb Pack is using ES to store the Transaction > > > related > > > > > > > > event[3]. I think we can provide a service to bridge the > Trace > > > > > > > > information that Skywalking collects with ServiceComb Pack > > > > > Transaction > > > > > > > > events information. > > > > > > > > > > > > > > > > Any thoughts? > > > > > > > > > > > > > > > > [1] > > > > > > > > > > > > > > > > > > > > > https://github.com/apache/servicecomb-pack/blob/master/omega/omega-context/src/main/java/org/apache/servicecomb/pack/omega/context/IdGenerator.java > > > > > > > > [2] > > > > > > > > > > > > > > > > > > > > > https://github.com/apache/servicecomb-pack/blob/master/omega/omega-context/src/main/java/org/apache/servicecomb/pack/omega/context/OmegaContext.java > > > > > > > > [3] > > > > > > > > > https://github.com/apache/servicecomb-pack/tree/master/alpha/alpha-fsm > > > > > > > > > > > > > > > > Willem Jiang > > > > > > > > > > > > > > > > Twitter: willemjiang > > > > > > > > Weibo: 姜宁willem > > > > > > > > > > > > > > > > > > > > > > > > > > > >
