Not all trace I can see in web ui.

[cosmonaut13]

I use zipkin4net and docker-zipkin to track traces in my application. I have a unit test. Inside a unit test a client is created from which I send a request to the first server, then I send the request from the first server to the second, then from the second service I send a request to the third service, and then I return the response to the client. All this I do with the help of commands ClientSend - ServerReciev-ServerSend-ClientReciev. If the request is one, then all is well. If I send asynchronously 100 requests, then some of the traces are lost. But if I debug the test in debug mode, or when I end the test, I set a delay, then all the traces are displayed. It seems that all the traces simply do not have time to send to the collector. Can you help me?

[fedj]

Hi @cosmonaut13,

Are you using the Rate Limiting system ? If so, it could explain that some spans are just discarded.

[cosmonaut13]

@fedj Hi. No, I don't use Rate Limiting test. For exmape:
-> start unit test
TraceManager.Start()
await (here send asynchronously 100 requests how Task.WhenAll(Enumerable(tasks)))
TraceManager.Stop()
->end unit test

[fedj]

Ok, can you also tell me the sampling rate you're using ?

[cosmonaut13]

@fedj sampling rate = 1.0

[cosmonaut13]

if i set Await Task.Delay() after TraceManager.Stop() but before end of unit test, i see all traces in web ui.

[fedj]

Ok. Can you try with a custom IRecordDispatcher ? The default one is async and could cause some issue (c.f.

zipkin4net/Src/zipkin4net/Src/TraceManager.cs

Line 68 in 7be090c

public static bool Start(ILogger logger, IRecordDispatcher dispatcher)

)

[cosmonaut13]

@fedj Do I understand correctly, I need to write my OrderAsyncDispatcher? I use standart IRecordDispatcher

[fedj]

@cosmonaut13 I understand you can be surprised, I'm just trying to understand where the async could come from and I think the easiest way would be to make it synchronous to check our assertions. Once the problem isolated, we will try to come up with a fix (probably trying to flush everything before stopping) but we need to verify assertions first.