Liu,Jiapeng <[email protected]> 于2020年1月2日周四 下午5:09写道：
> I have the following questions: > 1. how the span isError convert to the status of service scope, I have the > following questions > > 2. how the oal define in fllowing status: > In SkyWalking, it depends on the interface response of C, if C > propagates > the error because of b's response code, then yes, it will be recorded > as > one failure. > > C(method1 NORMAL )-> C(method2 EXCEPTION )-> B(method1 NORMAL > )->B(method2 EXCEPTION ) > It depends, is C#method1 an entry span(Tomcat or Spring MVC span) in SkyWalking? I am assuming yes. Then SLA will keep in 100%, because it responses to request in the healthy status, but the segment isError will be true. > > > > > > > 在 2020/1/2 下午4:20，“Sheng Wu”<[email protected]> 写入: > > Answer inline. > > > Liu,Jiapeng <[email protected]> 于2020年1月2日周四 下午4:06写道： > > > Hi All: > > Dear skywalking submitter,I am from Baidu Aifanfan,.We are currently > > selecting the tracing system.There are currently 2 alternatives > > 1. Skywalking > > 2. Baidu Internal System in feng chao > > Since the calculation of skywalking relies only on OAP clusters, > reducing > > the cost of deployment and operation and maintenance, currently it > is more > > inclined to choose skywalking > > As the current system sla calculation is relatively special, the > specific > > logic is as follows: > > One call across 3 systems.The call is c-> b-> a. If B is exception, > A is > > normal, C will also be recorded as an exception > > > > In SkyWalking, it depends on the interface response of C, if C > propagates > the error because of b's response code, then yes, it will be recorded > as > one failure. > > > > > > I want to ask All, service, entrypoint dimensions, what is the logic > of > > sla? At present our department will want to It is used in the test > > environment on k8s, but there is a problem in the calculation logic > of sla, > > because the relevant information is relatively small and cannot be > > determined at present, > > > > SLA today in SkyWalking is quite simple, SLA == successful rate, due > to the > OAL script. > > service_sla = from(Service.*).percent(status == true); > > You could override this logic by providing your own function at OAL > level. > > > > > The currently understood skywalking process: > > 1. java agent dot, generate span > > 2. Instance-level spans are aggregated into segments and issued > > 3. OAP parsing segments and local files > > 4. Stream processing > > > > > > Points in doubt (none found in the code): > > 1.The segment index in the es data has an isError field. Is this > part of > > the isError an OR relationship? If there is an error in a segment in > a > > segment, the overall segment is judged for error > > > > If there is at least one span in the segment tagged as an error, then > this > segment will be `isError=true`. > > > > 2.How does the isError of the spans in the segment converge to > statuses > > such as ALL, SERVICE, etc. > > > > That depends on the OAL script. Such as for service > > service_sla = from(Service.*).percent(status == true); > > You could find all in the official_analysis.oal in source code or > release > tar > > > > 3. If the access party's RBI interface throws a service exception, is > > there an ignored interface in skywalking? > > > > What is RBI interface? Today, all error is related to error status of > span > in the tracing scenario, > > https://github.com/apache/skywalking/blob/4e501328171f434c74144d373be6df0264f43a24/oap-server/server-receiver-plugin/skywalking-trace-receiver-plugin/src/main/java/org/apache/skywalking/oap/server/receiver/trace/provider/parser/listener/endpoint/MultiScopesSpanListener.java#L224 > . > How is span in error status, it depends on the plugin. Mostly, if an > exception or 400+ HTTP response code detected. > > Sheng Wu 吴晟 > Twitter, wusheng1108 > > > > > > I hope to get help from you, thankyou > > > > > > > > Evan.liu > > > > > > > > > > Evan.liu > > > > > > >
