[Design] Tracing story for LRO

[anuchandy]

This work item is to explore various options for LRO tracing.

The LRO abstraction quick intro:

The client-side LRO abstraction PollerFlux is one type that composes almost 4 service calls:

1. One time service call to start LRO
2. Multiple service call to poll LRO
3. Service call to retrieve the final result of LRO (if service support it)
4. Service call to cancel the LRO (if service support it)

Problem:

We need to decide on how we want to report telemetry for these service calls from a single PollerFlux.

--> One option is to put all these logically associated API calls under one group.

Examples:

The following examples show a couple of usage of PollerFlux and how telemetry looks like with grouping option.

Example1:

Start LRO and do the polls

fooFlux.map(response -> {
    Logger.log(response.getStatus());
    return response;
}).block();

Example2:

Start LRO, do the polls and fetch result.

fooFlux.flatmap(response -> {
    Logger.log(response.getStatus());
    if (response.getStatus() == LongRunningOperationStatus.SUCCEEDED)  {
    return response.getFinalResult();
} else {
    return Mono.empty();
}
}).block();

Example3:

Start LRO, do the polls then cancel.

fooFlux.flatmap(response -> {
    Logger.log(response.getStatus());
    if (response.getStatus() == LongRunningOperationStatus.BLOCKED)  {
    return response.cancelOperation();
} else {
    return Mono.empty();
}
}).block();

Example4:

Use the first subscription to Start LRO, do 2 polls

fooFlux.map(response -> {
    Logger.log(response.getStatus());
    return response;
}).take(2).block();

do subscription on the same poller flux instance for polling

fooFlux.map(response -> {
    Logger.log(response.getStatus());
    return response;
}).block();

Note one instance of PollerFlux makes an only single call to start LRO, so other subscription simply result in polling.

One way to show the telemetry is like below by grouping two subscription under the same parent span, if we decide this path then we could user parent PollerContext.

To discuss:

Is the above-outlined grouping option make sense for LRO, what are the other alternatives.

\cc @alzimmermsft @JonathanGiles @srnagar @samvaity

[github-actions]

Hi @anuchandy, we deeply appreciate your input into this project. Regrettably, this issue has remained inactive for over 2 years, leading us to the decision to close it. We've implemented this policy to maintain the relevance of our issue queue and facilitate easier navigation for new contributors. If you still believe this topic requires attention, please feel free to create a new issue, referencing this one. Thank you for your understanding and ongoing support.