`otel-collector` pause breaks the grpc `OTLPSpanExporter` permanently.

[cospectrum]

Describe your environment

OS: Darwin 22.6.0 Python version: 3.13.2 SDK version: v1.31.1 API version: v1.31.1 opentelemetry-exporter-otlp v1.31.1

What happened?

After otel-collector restart, the app failed to send traces and did not recover from the poisoned state. All subsequent grpc requests are broken.

Steps to Reproduce

1. Start the otel-collector (with grpc receiver):

services:
  otel-collector:
    image: otel/opentelemetry-collector-contrib:0.123.0
    ports:
      - 1888:1888 # pprof extension
      - 8888:8888 # Prometheus metrics exposed by the Collector
      - 8889:8889 # Prometheus exporter metrics
      - 13133:13133 # health_check extension
      - 4317:4317 # OTLP gRPC receiver
      - 4318:4318 # OTLP http receiver
      - 55679:55679 # zpages extension

2. Start your instrumented app to send traces via grpc. For example:

import os
import time

from opentelemetry import trace
from opentelemetry.exporter.otlp.proto.grpc.trace_exporter import OTLPSpanExporter
from opentelemetry.sdk.trace import TracerProvider
from opentelemetry.sdk.trace.export import BatchSpanProcessor
from opentelemetry.sdk.trace.id_generator import RandomIdGenerator
from opentelemetry.trace.propagation.tracecontext import TraceContextTextMapPropagator


def main() -> None:
    otel_collector_grpc_endpoint = "localhost:4317"
    ping_interval = 3

    trace_provider = TracerProvider()
    trace_provider.add_span_processor(
        BatchSpanProcessor(
            OTLPSpanExporter(
                endpoint=os.path.join(otel_collector_grpc_endpoint, "v1/traces"),
                insecure=True,
            )
        )
    )
    trace.set_tracer_provider(trace_provider)

    while True:
        traceparent = random_traceparent()
        ping(traceparent)
        time.sleep(ping_interval)


def ping(traceparent: str) -> None:
    tracer = trace.get_tracer(__name__)

    carrier = {"traceparent": traceparent}
    print(carrier)
    ctx = TraceContextTextMapPropagator().extract(carrier)

    with tracer.start_as_current_span("ping", ctx):
        pass


def random_traceparent() -> str:
    to_hex = lambda s: hex(s)[2:]
    gen = RandomIdGenerator()
    trace_id = to_hex(gen.generate_trace_id()).zfill(32)
    span_id = to_hex(gen.generate_span_id()).zfill(16)
    return f"00-{trace_id}-{span_id}-01"


if __name__ == "__main__":
    main()

3. Pause the otel-collector for more than ping_interval.
4. Resume the otel-collector.
5. Check that app fails to send new traces.

Expected Result

The instrumented app should send traces successfully after otel-collector restart.

Actual Result

The app fails to send new traces after otel-collector resumed.

Additional context

After otel-collector restart, the python app gives logs like that (for all batches):

Transient error StatusCode.UNAVAILABLE encountered while exporting traces to localhost:4317/v1/traces, retrying in 8s.

The only way to fix the app tracing is to restart the app itself, which is painful. Also http mode OTLPSpanExporter seems to be working correctly unlike grpc.

Would you like to implement a fix?

None

[xrmx]

Dup of #4517 ?

[cospectrum]

Looks like it, but I encountered a problem with grpc traces instead of metrics. Maybe there is a problem with _logs too.

[devmonkey22]

Looks like it, but I encountered a problem with grpc traces instead of metrics. Maybe there is a problem with _logs too.

All 3 signals (metrics, logs, traces) never recovered with this scenario on my end too.