Otel optimizations I

[brunobat]

During the otel performance work some inefficiencies were detected on the hot path, the one affecting the creation of all spans:

1. We were placing MDC trace data in a HashMap that was resizing. Code was re-written in order to not use the HashMap at all.
2. When calculating host and port we where using String.split() which uses a regex and was costing many cycles. Re-wrote that part and added tests to make sure behaviour is compatible.

[quarkus-bot]

/cc @radcortez (opentelemetry)

[quarkus-bot]

Status for workflow Quarkus CI

This is the status report for running Quarkus CI on commit c65d7ac.

✅ The latest workflow run for the pull request has completed successfully.

It should be safe to merge provided you have a look at the other checks in the summary.

You can consult the Develocity build scans.

[geoand]

LGTM

[brunobat]

FYI, After profiling, the test run used 4% less samples.

[gsmet]

I wouldn't add these comments if you weren't already into micro-optimization mode but:

• if span can be null quite often, you could also return Collections.emptyMap() in this case
• I also wonder if it might be worth using Map.of and specializing the cases a bit more

I was thinking of something like:

    public static Map<String, String> getSpanData(Context context) {
        if (context == null) {
            return Collections.emptyMap();
        }

        Span span = Span.fromContextOrNull(context);
        if (span == null) {
            return Collections.emptyMap();
        }

        SpanContext spanContext = span.getSpanContext();
        if (span instanceof ReadableSpan readableSpan
                && readableSpan.getParentSpanContext() != null
                && readableSpan.getParentSpanContext().isValid()) {
            return Map.of(
                    SPAN_ID, spanContext.getSpanId(),
                    TRACE_ID, spanContext.getTraceId(),
                    SAMPLED, Boolean.toString(spanContext.isSampled()),
                    PARENT_ID, readableSpan.getParentSpanContext().getSpanId()
            );
        }

        return Map.of(
                SPAN_ID, spanContext.getSpanId(),
                TRACE_ID, spanContext.getTraceId(),
                SAMPLED, Boolean.toString(spanContext.isSampled())
        );
    }

It might not be worth the additional trouble though so I will let you be the judge of if. I don't expect this code to change often so I think the duplication is not that big of a problem if it performs faster.

Also the second part only works if we expect the Map to be immutable, obviously :).

[brunobat]

Immutability is why I didn't think about the Map.of()... I'll give it a try.
Please mind the execution path using the HashMap is now only used for logging.

[gsmet]

Yeah, if it doesn't look like a good idea, just don't bother :).
The first part about not creating the Map if null might be a good idea though.

I only added a comment because it was already micro optimization but at some point you have to decide that things are good enough!