[OTEL 0.22] ease the shutdown of SdkMeterProvider?

[Oliboy50]

Why talking about an opentelemetry update in tracing-opentelemetry?

I already opened a discussion in open-telemetry/opentelemetry-rust, then someone there told me that I should open a discussion here instead.

What is the issue?

My issue is about how to properly shutdown the SdkMeterProvider since there is no more opentelemetry::global::shutdown_meter_provider() in opentelemetry 0.22.

Here is an example of what it can look like to build a simple wrapper library over OpenTelemetry:

pub fn build_metrics_layer() -> tracing_opentelemetry::MetricsLayer<S>
    where
        S: tracing::Subscriber + for<'span> tracing_subscriber::registry::LookupSpan<'span>,
{
    let meter_provider = opentelemetry_otlp::new_pipeline()
        .metrics(opentelemetry_sdk::runtime::Tokio)
        .with_exporter(some_exporter())
        .build()
        .expect("Failed to build OTLP meter provider");

    MetricsLayer::new(meter_provider)
}

pub fn shutdown_meter_provider() {
    // this work in opentelemetry 0.21, 
    // but in opentelemetry 0.22, this feature is gone! :'(
    opentelemetry::global::shutdown_meter_provider();
}

Then we would use this lib in a project, such as:

let registry = tracing_subscriber::registry()
    .with(tracing_subscriber::filter::EnvFilter::from_default_env())
    .with(build_metrics_layer());
tracing::subscriber::set_global_default(registry).expect("Failed to set global tracing subscribers");

// do our stuff, such as run an HTTP API server
// ...

shutdown_meter_provider();

So how are we supposed to shutdown the SdkMeterProvider used internally, without using opentelemetry::global::shutdown_meter_provider()? 🙏

Currently, I found a workaround which is to keep my own static instance of the SdkMeterProvider so that I can shutdown it later, such as:

use std::sync::OnceLock;
use opentelemetry_sdk::metrics::SdkMeterProvider;

static METER_PROVIDER: OnceLock<SdkMeterProvider> = OnceLock::new();

pub fn build_metrics_layer() -> tracing_opentelemetry::MetricsLayer<S>
    where
        S: tracing::Subscriber + for<'span> tracing_subscriber::registry::LookupSpan<'span>,
{
    let meter_provider = opentelemetry_otlp::new_pipeline()
        .metrics(opentelemetry_sdk::runtime::Tokio)
        .with_exporter(some_exporter())
        .build()
        .expect("Failed to build OTLP meter provider");

    METER_PROVIDER.set(meter_provider.clone()).expect(
        "build_metrics_layer() is not meant to be used more than once in the same process",
    );

    MetricsLayer::new(meter_provider)
}

pub fn shutdown_meter_provider() {
    if let Some(meter_provider) = METER_PROVIDER.get() {
        meter_provider
            .shutdown()
            .expect("Failed to shutdown meter provider");
    }
}

This workaround seems to work fine, however I don't know if this is a good practice...
WDYT?

Do you understand why lalitb told me to open the discussion on the tracing-opentelemetry lib instead of the opentelemetry libs?

[Oliboy50]

Closing this issue since this has been fixed on the OpenTelemetry side in this PR

Update: see answer below

[cijothomas]

https://github.com/open-telemetry/opentelemetry-rust/pull/1739/files See this, which shows the right way of shutdown. There won't be global shutdown method. Apologies for going back and forth on this issue.