opentracing/public

All things OpenTracing

Michael Timberg

@michaeltimberg

Some context for ☝️: added the opentracing agent instead of the opentelemery agent to a react frontend (communicating with a dotnet backend with an otel tracer).

need to determine if switching to opentelemetry and redoing auto. and manual instrumentation would be easier than trying to get opentracing agent to always use w3c headers.

Richard Fine

@richard-fine

What's the current state of spec/RFCs/ecosystem/etc with regards to sequencing and 'happens after' relationships between spans? I don't mean FollowsFrom (which is causal), I mean something temporal. I tried to restart conversation about this back in 2019 (in opentracing/specification#142) but didn't get any takers. Has anything moved onwards?

Richard Fine

@richard-fine

(I guess maybe I should ask OpenTelemetry instead)

Yuri Shkuro

@yurishkuro

+1 OpenTelemetry. I do not expect any changes to OpenTracing spec going forward.

Ilya Shipitsin

@chipitsine

how do you install it on Ubuntu ? is there well known repo ? particularly I'm interested in cpp

Андрей Калинин

@kalinin81

Hi. Is there the “best practice” for “online” switch (without restarting traced module) “on/off” tracing?

Mukund B Srinivas

@mukundsrinivasb

Hello world,
I am new to opentracing , I speak java decently well and C++ not so fluently. I am looking for some good reading and coding material to get me up to speed on some good first issues .
Thanks a lot.

Nhung

@Nhung22461767_twitter

Hi everyone, thanks for being here!
I'm using latest Quarkus versions (2.0.0.Final and quarkus-smallrye-opentracing 2.0.0.Final) it looks like after calling mongo with reactive operators trace context is resetted (ie one cannot get back traceId). If mongo is not called it works as expected.
It looks like this:
tracer.spanId( ) // has a value
.flatMap { repo.insert(person) ) }
.flatMap { tracer.spanId( ) } // value is null
Do you know of any solutions to keep the trace context after calling mongodb with reactive operators?

lteqgt

@sebastianharko

Hi, I love OpenTracing and I'm using the Java client to add tracing to a Scala/Akka application .... The question I have is how do I get a Span from a SpanContext, I'm using tracer.extract to extract from a hash map (Map<String, String>...) but all I'm getting is a SpanContext ... and the question how do i get the 'original' Span so I can finish it (so the span started on a different node and I want to finish it on this node so I can track the 'network' travel time ...

Yuri Shkuro

@yurishkuro

@sebastianharko spans are meant to belong to a single process, not represent the time range between two processes

you can use span context to start a new child span

lteqgt

@sebastianharko

@yurishkuro thanks for the answer, that makes total sense, so how do people track 'network time' (time it takes for a message to travel across the network between two processes) ?

Yuri Shkuro

@yurishkuro

@sebastianharko when the trace is assembled on the backend you would have that information, e.g. as the difference in start time of producer and consumer spans

lteqgt

@sebastianharko

@yurishkuro so if the span on the consumer side is extracted as a "child" of the span on the producer side .....then in something like ZipKin or OpenTracing, the network time would show up as the 'white space' between the two spans, right ?

Yuri Shkuro

@yurishkuro

@sebastianharko yes

Quoc-Anh Nguyen

@imcvampire

I would like to contribute to this repo - https://github.com/opentracing-contrib/goredis - to support the latest version of goredis - v8: opentracing-contrib/goredis#6

Which process should I take?

Ankit Mehta

@mehta-ankit_gitlab

Can we get someone to release https://github.com/opentracing-contrib/java-jdbc. We made a PR and it got merged but we are waiting for a new release.

It was merged 9 days ago.

Henrique Luis Schmidt

@henriquels25

Hi, I have a problem with https://github.com/opentracing-contrib/java-spring-cloud. I can't use the latest version of Spring Boot and Spring Cloud OpenFeign because of classpath problems.

The PR opentracing-contrib/java-spring-cloud#324 would solve the problem but it is open since Apr 26.

Could someone look at it and merge the PR if everything is OK?

Ayushi Jain

@ayushi19031

Hello, I am new to open source. Is there anything I can start working on?

Yuri Shkuro

@yurishkuro

@ayushi19031 please check the OpenTelemetry project instead, all activities have moved there.

Erkan Erol

@erkanerol

Hi. Is there a way to serialize/deserialize SpanContext in Go? There are some examples for HTTP and RPC but I couldn't find how to transport SpanContext for customization.

Yuri Shkuro

@yurishkuro

define a carrier that is just a map, and use Inject to serialize into that carrier

@erkanerol ^

Erkan Erol

@erkanerol

@yurishkuro Just got it now.

In parent side, the code below gets a map with key "uber-trace-id" for Jaeger.

textCarrier := opentracing.TextMapCarrier{}
err := span.Tracer().Inject(span.Context(), opentracing.TextMap, textCarrier)

Then I can get parent context like

textCarrier := opentracing.TextMapCarrier{"uber-trace-id": "2975a5cc24621e8a:2975a5cc24621e8a:0000000000000000:1"}
spanCtx, err := tracer.Extract(opentracing.TextMap, textCarrier)

All I need to transport that encoded string. Nice :)

Erkan Erol

@erkanerol

Now I have another question: Is there a way to start/finish a span without keeping it in memory and by re-creating it again with some references to finish? For example, a k8s controller doesn't keep anything in memory for a CustomResource. It fetches everything from api-server again and again. I want to start a span when I observed a new CR and then finish it after a while depending on some status fields. Meanwhile, there is no state on k8s operator so that I can store the span object. Therefore, I need to regenerate it from scratch to finish. Is there a way to do it?

Erkan Erol

@erkanerol

In "https://opentracing.io/guides/golang/spans/" under "Passing of scopes to another thread or callback is not supported.", there is such a statement:

Passing of scopes to another thread or callback is not supported.

Does it mean that answer of my question is "It is not possible. "?

Mathias Herberts

@hbs

Hi, what is the status of PIN 14 and the availability of event driven flows? What is the most current pattern used to support event-driven like executions?

Francisco Cha

@fcha_gitlab

Hi, I was attempting to access the span in Java defined by the documentation https://opentracing.io/guides/java/spans. However my version opentracing-util-0.22.0 the object Tracer doesn't have a scopeManager().

Rodrigo Bento

@rbento

Hello,
I am getting started with distributed tracing. Should I invest any time on OpenTracing or start with OpenTelemetry?

2 replies

Also, if I have a project currently using the OpenTracing API, is it advisable to migrate it to the OpenTelemetry API?

Thank you

Raphael Deem

@r0fls

I asked this over the weekend and deleted it, but I'm still stuck and don't know where else to go for help...

Basically I am seeing these spans with the incorrect duration, they show what appears to be the right start time, but they have the duration of the complete web request, which doesn't even make sense.

Screen Shot 2021-10-25 at 12.32.38 PM.png

I'm trying to adapt this gql tracing to our needs: https://github.com/99designs/gqlgen-contrib/blob/master/gqlopentracing/tracer.go

It is the operation traces that are showing the incorrect duration. I'm just looking for troubleshooting tips, since going into the golang debugger isn't really helping me much... how can the shild spans have a duration such that the duration + start time is greater than the whole request duration?

This is why I dislike the golang "copying is better than deps" mantra: normally I would raise this issue on that repo, but the whole thing is neglected as it seems the community has just been copying it rather than using it as a dep as indicated by this thread https://github.com/99designs/gqlgen-contrib/issues/26#issuecomment-946907900

Raphael Deem

@r0fls

still seeing really weird duration/start times combos that don't make any sense at all on my spans originating from gqlgen... if anyone has any experience with debugging this type of thing, I will venmo you