Parent and child spans timing

[juliaabrz]

Hi, I'm creating a flamegraph tool for traces from Jaeger and was wondering about the timing of spans. Is it always true that the duration of the children span is always contained within a parent span, or are there cases where the child takes longer than the parent did (and thus the parent does not need to wait for the child to be finished in order to be complete)? I appreciate the help explaining this concept to me.

[dosubot[bot]]

Hello @juliaabrz! I'm here to help you with any questions or issues you have. Let's tackle this one together.

No, it is not always true that the duration of the child spans is contained within a parent span. There are cases where the child span takes longer than the parent span. For example, in Jaeger, there is a test case that demonstrates a scenario where a single child span is longer than its parent span:

it('span with single child span longer than its parent', () => {
    const [serviceOne] = getColumnValues(
      'Service Name',
      transformedtraceWithSingleChildLongerThanParentAndStartsAfterParent
    );
    expect(serviceOne.selfTotal).toBe(0.04);
});

This shows that the parent does not need to wait for the child to be finished in order to be complete ^[1].

To continue talking to Dosu, mention @dosu.

[juliaabrz]

@dosu where can that test case be used? Are you able to explain the difference between a callstack and a trace?

[dosubot[bot]]

The test case for a span with a single child span longer than its parent can be used in the context of validating the duration and timestamp fields of spans in a trace. This ensures that the tracing system correctly handles and records spans where a child span's duration exceeds that of its parent span.

The difference between a call stack and a trace is as follows:

• Call Stack: A call stack is a data structure that stores information about the active subroutines or functions of a computer program. It is used to keep track of function calls and their return addresses, local variables, and other state information. The call stack is specific to a single thread of execution and represents the sequence of function calls that have been made up to a certain point in time.

• Trace: A trace, in the context of distributed tracing, is a series of spans that represent a single transaction or request as it flows through various services and components in a distributed system. Each span represents a single unit of work or operation, such as an RPC call, and includes metadata such as timestamps, duration, and relationships to other spans (e.g., parent-child relationships). A trace provides a holistic view of the entire transaction, showing how it propagates through the system and where time is spent ^[1].

To continue talking to Dosu, mention @dosu.

[yurishkuro]

@juliaabrz note that Jaeger UI already shows flamegraphs