Update default logic for OTel top-level spans identification

[liustanley]

What does this PR do?

Updates default logic to improve top-level spans identification in OTLP ingest. Users have the option to disable this new logic if the old logic is preferred by using the APM Feature "disable_otlp_compute_top_level_by_span_kind" The new logic is as follows for OTLP spans:

• Root spans and spans with a server or consumer span.kind will be marked as top-level.
• Additionally, spans with a client or producer span.kind will have stats computed (marked as measured).

Also adds a telemetry metric datadog.trace_agent.otlp.compute_top_level_by_span_kind in order to track performance of this feature in beta.

Motivation

See RFC.

Additional Notes

Ran benchmark tests and verified that there are no notable changes in performance.

Main benchmark (control):

❯ go test -run=XXX -bench=BenchmarkProcessRequest -tags=test
goos: darwin
goarch: arm64
pkg: github.com/DataDog/datadog-agent/pkg/trace/api
BenchmarkProcessRequest-10    	     141	   8413292 ns/op	10020732 B/op	    8313 allocs/op
PASS
ok  	github.com/DataDog/datadog-agent/pkg/trace/api	2.674s

PR benchmark:

❯ go test -run=XXX -bench=BenchmarkProcessRequest -tags=test
goos: darwin
goarch: arm64
pkg: github.com/DataDog/datadog-agent/pkg/trace/api
BenchmarkProcessRequest-10    	     139	   8557711 ns/op	10228702 B/op	    9312 allocs/op
PASS
ok  	github.com/DataDog/datadog-agent/pkg/trace/api	2.692s

Possible Drawbacks / Trade-offs

Describe how to test/QA your changes

Send OTLP spans of varying span kinds and verify that root spans and server/consumer spans are marked as top-level in Datadog. Also verify that client/producer spans are marked as measured and have stats computed, and internal spans are not marked as top-level or measured.

Finally, verify that adding the APM feature flag "disable_otlp_compute_top_level_by_span_kind" reverts to the old top-level spans logic.

Reviewer's Checklist

• If known, an appropriate milestone has been selected; otherwise the Triage milestone is set.
• Use the major_change label if your change either has a major impact on the code base, is impacting multiple teams or is changing important well-established internals of the Agent. This label will be use during QA to make sure each team pay extra attention to the changed behavior. For any customer facing change use a releasenote.
• A release note has been added or the changelog/no-changelog label has been applied.
• Changed code has automated tests for its functionality.
• Adequate QA/testing plan information is provided. Except if the qa/skip-qa label, with required either qa/done or qa/no-code-change labels, are applied.
• At least one team/.. label has been applied, indicating the team(s) that should QA this change.
• If applicable, docs team has been notified or an issue has been opened on the documentation repo.
• If applicable, the need-change/operator and need-change/helm labels have been applied.
• If applicable, the k8s/<min-version> label, indicating the lowest Kubernetes version compatible with this feature.
• If applicable, the config template has been updated.

[pr-commenter]

Bloop Bleep... Dogbot Here

Regression Detector Results

Run ID: 28defe4a-bbe8-426a-b81f-dccebb2f89be
Baseline: 8c5ec0f
Comparison: 1213406

Performance changes are noted in the perf column of each table:

• ✅ = significantly better comparison variant performance
• ❌ = significantly worse comparison variant performance
• ➖ = no significant change in performance

No significant changes in experiment optimization goals

Confidence level: 90.00%
Effect size tolerance: |Δ mean %| ≥ 5.00%

There were no significant changes in experiment optimization goals at this confidence level and effect size tolerance.

Experiments ignored for regressions

Regressions in experiments with settings containing erratic: true are ignored.

perf	experiment	goal	Δ mean %	Δ mean % CI
➖	file_to_blackhole	% cpu utilization	+0.46	[-6.33, +7.26]

Fine details of change detection per experiment

perf	experiment	goal	Δ mean %	Δ mean % CI
➖	otel_to_otel_logs	ingress throughput	+0.47	[-0.16, +1.09]
➖	file_to_blackhole	% cpu utilization	+0.46	[-6.33, +7.26]
➖	trace_agent_json	ingress throughput	+0.01	[-0.02, +0.03]
➖	trace_agent_msgpack	ingress throughput	+0.00	[-0.01, +0.01]
➖	tcp_dd_logs_filter_exclude	ingress throughput	+0.00	[-0.00, +0.00]
➖	uds_dogstatsd_to_api	ingress throughput	-0.00	[-0.00, +0.00]
➖	process_agent_standard_check_with_stats	memory utilization	-0.04	[-0.07, -0.01]
➖	process_agent_standard_check	memory utilization	-0.08	[-0.11, -0.04]
➖	file_tree	memory utilization	-0.49	[-0.57, -0.41]
➖	process_agent_real_time_mode	memory utilization	-0.73	[-0.77, -0.68]
➖	idle	memory utilization	-0.73	[-0.77, -0.68]
➖	basic_py_check	% cpu utilization	-0.90	[-3.15, +1.35]
➖	tcp_syslog_to_blackhole	ingress throughput	-0.92	[-0.97, -0.87]
➖	uds_dogstatsd_to_api_cpu	% cpu utilization	-1.33	[-2.75, +0.10]

Explanation

A regression test is an A/B test of target performance in a repeatable rig, where "performance" is measured as "comparison variant minus baseline variant" for an optimization goal (e.g., ingress throughput). Due to intrinsic variability in measuring that goal, we can only estimate its mean value for each experiment; we report uncertainty in that value as a 90.00% confidence interval denoted "Δ mean % CI".

For each experiment, we decide whether a change in performance is a "regression" -- a change worth investigating further -- if all of the following criteria are true:

1. Its estimated |Δ mean %| ≥ 5.00%, indicating the change is big enough to merit a closer look.

2. Its 90.00% confidence interval "Δ mean % CI" does not contain zero, indicating that if our statistical model is accurate, there is at least a 90.00% chance there is a difference in performance between baseline and comparison variants.

3. Its configuration does not mark it "erratic".

[ajgajg1134]

A few comments and questions here, also are there any integration or e2e tests that could be added here? I saw unit tests covering some of the changes, but given the importance of top-level and measured spans I think this change warrants some larger scoped tests to verify the full pipeline with configuration works as expected

[brett0000FF]

👍 from Docs with minor suggestion.

[ajgajg1134]

Also the Stats Concentrator in the trace-agent already has "computeStatsForSpanKind" (configured via apm_config.compute_stats_by_span_kind) that seems to overlap with the functionality here, is this intentionally looking to replace that?

[liustanley]

@ajgajg1134

A few comments and questions here, also are there any integration or e2e tests that could be added here? I saw unit tests covering some of the changes, but given the importance of top-level and measured spans I think this change warrants some larger scoped tests to verify the full pipeline with configuration works as expected

Agreed, I can look into adding system tests but that would be separate from this PR.

Also the Stats Concentrator in the trace-agent already has "computeStatsForSpanKind" (configured via apm_config.compute_stats_by_span_kind) that seems to overlap with the functionality here, is this intentionally looking to replace that?

Yes this is meant to replace that for OTel spans, the new functionality has the same behavior as computeStatsForSpanKind with the addition of top-level spans identification. computeStatsForSpanKind is only used here in the code, so the new logic would effectively override this option if both are set.

[pr-commenter]

Test changes on VM

Use this command from test-infra-definitions to manually test this PR changes on a VM:

inv create-vm --pipeline-id=30374733 --os-family=ubuntu

[pr-commenter]

Regression Detector

Regression Detector Results

Run ID: 04767d10-0f8d-4a6a-9736-d3e303fddc57
Baseline: a670800
Comparison: d3ed562

Performance changes are noted in the perf column of each table:

• ✅ = significantly better comparison variant performance
• ❌ = significantly worse comparison variant performance
• ➖ = no significant change in performance

No significant changes in experiment optimization goals

Confidence level: 90.00%
Effect size tolerance: |Δ mean %| ≥ 5.00%

There were no significant changes in experiment optimization goals at this confidence level and effect size tolerance.

Experiments ignored for regressions

Regressions in experiments with settings containing erratic: true are ignored.

perf	experiment	goal	Δ mean %	Δ mean % CI
➖	file_to_blackhole	% cpu utilization	+1.37	[-4.98, +7.71]

Fine details of change detection per experiment

perf	experiment	goal	Δ mean %	Δ mean % CI
➖	file_to_blackhole	% cpu utilization	+1.37	[-4.98, +7.71]
➖	tcp_syslog_to_blackhole	ingress throughput	+0.15	[+0.07, +0.24]
➖	process_agent_real_time_mode	memory utilization	+0.07	[+0.04, +0.11]
➖	trace_agent_msgpack	ingress throughput	+0.03	[+0.02, +0.04]
➖	trace_agent_json	ingress throughput	+0.01	[-0.02, +0.03]
➖	uds_dogstatsd_to_api	ingress throughput	+0.00	[-0.20, +0.20]
➖	pycheck_1000_100byte_tags	% cpu utilization	-0.00	[-4.87, +4.87]
➖	tcp_dd_logs_filter_exclude	ingress throughput	-0.01	[-0.04, +0.01]
➖	process_agent_standard_check_with_stats	memory utilization	-0.10	[-0.13, -0.07]
➖	process_agent_standard_check	memory utilization	-0.19	[-0.22, -0.16]
➖	idle	memory utilization	-0.20	[-0.24, -0.17]
➖	basic_py_check	% cpu utilization	-0.28	[-2.66, +2.11]
➖	uds_dogstatsd_to_api_cpu	% cpu utilization	-0.30	[-2.97, +2.36]
➖	file_tree	memory utilization	-0.33	[-0.42, -0.25]
➖	otel_to_otel_logs	ingress throughput	-0.35	[-0.75, +0.05]

Explanation

A regression test is an A/B test of target performance in a repeatable rig, where "performance" is measured as "comparison variant minus baseline variant" for an optimization goal (e.g., ingress throughput). Due to intrinsic variability in measuring that goal, we can only estimate its mean value for each experiment; we report uncertainty in that value as a 90.00% confidence interval denoted "Δ mean % CI".

For each experiment, we decide whether a change in performance is a "regression" -- a change worth investigating further -- if all of the following criteria are true:

1. Its estimated |Δ mean %| ≥ 5.00%, indicating the change is big enough to merit a closer look.

2. Its 90.00% confidence interval "Δ mean % CI" does not contain zero, indicating that if our statistical model is accurate, there is at least a 90.00% chance there is a difference in performance between baseline and comparison variants.

3. Its configuration does not mark it "erratic".

[ajgajg1134]

Looks good from APM!

[brett0000FF]

Looks good from Docs.

[liustanley]

/merge

[dd-devflow]

🚂 MergeQueue

This merge request is not mergeable yet, because of pending checks/missing approvals. It will be added to the queue as soon as checks pass and/or get approvals.
Note: if you pushed new commits since the last approval, you may need additional approval.
You can remove it from the waiting list with /remove command.

Use /merge -c to cancel this operation!

[dd-devflow]

🚂 MergeQueue

Added to the queue.

There are 2 builds ahead of this PR! (estimated merge in less than 27m)

Use /merge -c to cancel this operation!