Original exception not displayed in log due to java.lang.IllegalStateException: Recording already ended

[dcdh]

Describe the bug

I have got a quarkus application using opentelemetry extension pushing span using vertx grpc.

Spans record a lot of stacktrace due to a technical exception. Because this is so huge, vertx grpc throw an io.vertx.grpc.common.MessageSizeOverflowException when pushing the batch spans.

The stack in logs return

Uncaught exception received by Vert.x: java.lang.IllegalStateException : recording already ended

Steps to reproduce

I do not have a public reproducer.

What did you expect to see?

I expect to see the original exception from grpc (MessageSizeOverflowException)
Also, maybe the content too, to beter understand the cause of the exception.

What did you see instead?

Only a stacktrace about recording already ended.

What version and what artifacts are you using?

quarkus 3.31.4
opentelemetry-sdk: 1.57.0

Environment
Compiler: openjdk version "25" Temurin-25+36 (build 25+36-LTS)
OS: windows 11

Additional context

I guess a way to fix it is to log before recording a failure.

In GrpcExporter, on the method onError by changing the location of metricRecording.finishFailed(e) after calls to logger and before doing result.failExceptionally(... should display the original exception.

_{Tip: React with 👍 to help prioritize this issue. Please use comments to provide useful context, avoiding +1 or me too, to help us triage it. Learn more here.}

[jkwatson]

I don't think that we maintain the vertx grpc exporter, so you'll need to log a bug with whomever maintains that library.

[trask]

cc @brunobat

[brunobat]

@dcdh , this should have been filed against https://github.com/quarkusio/quarkus/issues

That exception does indeed happen when you try to send gRPC frames that are too big. There are a couple of things you can do to mitigate the issues.

Limit the attribute size by setting (I would set this one):
quarkus.otel.attribute.value.length.limit

If you really must, you can also increase the size of a gRPC frame:
quarkus.grpc.clients."client-name".max-inbound-message-size

For the underlying condition I need more concrete pointers in order to help. When you mention:

In GrpcExporter, on the method onError by changing the location of metricRecording.finishFailed(e) after calls to logger and before doing result.failExceptionally(...

Can you please add links to the actual code?

[jkwatson]

I'm going to close this; any further discussion should happen where @brunobat mentioned.

[dcdh]

On my side I will try to do a reproducer. I cannot share the current code. Many thanks.