feat(opentelemetry source): support trace ingestion

[caibirdme]

This PR supports ingesting trace data in opentelemetry-protocol

Here's the example config:

sources:
  foo:
    type: opentelemetry
    grpc:
      address: 0.0.0.0:4317
    http:
      address: 0.0.0.0:4318
      keepalive:
        max_connection_age_jitter_factor: 0.1
        max_connection_age_secs: 300
sinks:
  bar:
    type: console
    inputs: [foo.traces]
    encoding:
      codec: json
  baz:
    type: console
    inputs: [foo.logs]
    encoding:
      codec: logfmt

we can use opentelemetry-trace-gen to test it.

1. install: go install github.com/open-telemetry/opentelemetry-collector-contrib/cmd/telemetrygen@latest
2. produce some demo traces: telemetrygen traces --otlp-insecure --duration 30s
3. produce some demo logs: telemetrygen logs --duration 30s --otlp-insecure

[gcuberes]

@caibirdme, by any chance are you also working on an opentelemetry sink for traces?

[caibirdme]

@caibirdme, by any chance are you also working on an opentelemetry sink for traces?

@gcuberes In technical terms, the issue of converting data to the OpenTelemetry (OTel) format should not be difficult. The process only involves serializing the data into a specified Protocol Buffers (PB) format and sending it via gRPC or HTTP to a fixed endpoint. However, there are some challenges in terms of how to use this approach.

Firstly, users can modify the received trace data through the VRL, which makes it difficult to ensure that the final sent data still conforms to the OTel protocol specifications (some fields may be missing or have incorrect types). Currently, I have not found any method to constrain the schema of sink data.

Secondly, if we want to convert any logs into OTel traces, users need to configure all the required fields for OTel traces, which can be cumbersome in terms of usage.

If the goal is simply to receive OpenTelemetry traces and then forward them unchanged to another location using the OTel protocol, similar to a proxy, this may be an acceptable solution for some use cases but I don't know if this is what people want.

In summary, while technically feasible, there are some challenges in terms of ensuring data conformance to the OTel protocol and ease of use for the end-users.

[caibirdme]

Anybody take a look on this pr?

[StephenWakely]

Hi, sorry this one slipped my attention. Vector has pretty poor support for traces, the only sink that can send out traces is the Datadog sink, but even then the format of those traces needs to be pretty precise. Anything else would be fairly error prone. I'm curious what your use case for this source would be?

[caibirdme]

We're using clickhouse to store the logs&traces, so clickhouse sink is enough for us. The path looks like client -> send otlp trace/log -> vector -> vrl -> clickhouse. And we want to try loki in the future, the loki sink is also supported

[jszwedko]

We're using clickhouse to store the logs&traces, so clickhouse sink is enough for us. The path looks like client -> send otlp trace/log -> vector -> vrl -> clickhouse. And we want to try loki in the future, the loki sink is also supported

Gotcha. So you are converting the trace to a log event to send it to one of those sinks? I'm ok with us adding this, it's just that the user experience will be pretty rough so I'd want to label it as experimental in the docs. Do you mind making the docs updates by adding a new "how it works" section here:

vector/website/cue/reference/components/sources/opentelemetry.cue

Lines 197 to 207 in 58a4a2e

	how_it_works: {
	tls: {
	title: "Transport Layer Security (TLS)"
	body: """
	Vector uses [OpenSSL](\(urls.openssl)) for TLS protocols due to OpenSSL's maturity. You can
	enable and adjust TLS behavior via the `grpc.tls.*` and `http.tls.*` options and/or via an
	[OpenSSL configuration file](\(urls.openssl_conf)). The file location defaults to
	`/usr/local/ssl/openssl.cnf` or can be specified with the `OPENSSL_CONF` environment variable.
	"""
	}
	}

?

[caibirdme]

Gotcha. So you are converting the trace to a log event to send it to one of those sinks? I'm ok with us adding this, it's just that the user experience will be pretty rough so I'd want to label it as experimental in the docs. Do you mind making the docs updates by adding a new "how it works" section here:

Sure, but I don't know exactly what should be to put into how it works. You mean how the source ingests trace data and converts it into log event?

[jszwedko]

Gotcha. So you are converting the trace to a log event to send it to one of those sinks? I'm ok with us adding this, it's just that the user experience will be pretty rough so I'd want to label it as experimental in the docs. Do you mind making the docs updates by adding a new "how it works" section here:

Sure, but I don't know exactly what should be to put into how it works. You mean how the source ingests trace data and converts it into log event?

Right, I would just add something like about it being experimental, have limited processing functionality available, and that the internal data format may change in the future.

[StephenWakely]

Nice. The code is looking pretty solid, just a fairly minor suggestion. Can you add an integration tests that covers this here.

This will also need a changelog entry. Can you add one here.

[github-actions]

Regression Detector Results

Run ID: 71d27108-0499-4b38-b646-71f9e463d748
Baseline: 234b126
Comparison: 4a31e4e
Total CPUs: 7

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

Significant changes in experiment optimization goals

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

perf	experiment	goal	Δ mean %	Δ mean % CI
❌	syslog_log2metric_tag_cardinality_limit_blackhole	ingress throughput	-6.06	[-6.18, -5.94]

Fine details of change detection per experiment

perf	experiment	goal	Δ mean %	Δ mean % CI
➖	otlp_http_to_blackhole	ingress throughput	+4.49	[+4.38, +4.61]
➖	otlp_grpc_to_blackhole	ingress throughput	+2.02	[+1.92, +2.11]
➖	datadog_agent_remap_blackhole	ingress throughput	+1.61	[+1.52, +1.71]
➖	socket_to_socket_blackhole	ingress throughput	+0.76	[+0.70, +0.83]
➖	http_text_to_http_json	ingress throughput	+0.51	[+0.39, +0.63]
➖	fluent_elasticsearch	ingress throughput	+0.30	[-0.18, +0.78]
➖	http_to_s3	ingress throughput	+0.21	[-0.07, +0.49]
➖	datadog_agent_remap_datadog_logs_acks	ingress throughput	+0.19	[+0.11, +0.27]
➖	http_to_http_acks	ingress throughput	+0.18	[-1.18, +1.54]
➖	http_to_http_noack	ingress throughput	+0.10	[+0.02, +0.18]
➖	http_to_http_json	ingress throughput	+0.03	[-0.05, +0.10]
➖	splunk_hec_indexer_ack_blackhole	ingress throughput	-0.00	[-0.15, +0.14]
➖	splunk_hec_to_splunk_hec_logs_acks	ingress throughput	-0.01	[-0.14, +0.13]
➖	splunk_hec_to_splunk_hec_logs_noack	ingress throughput	-0.06	[-0.17, +0.05]
➖	enterprise_http_to_http	ingress throughput	-0.07	[-0.15, +0.02]
➖	http_elasticsearch	ingress throughput	-0.53	[-0.60, -0.45]
➖	syslog_log2metric_humio_metrics	ingress throughput	-0.73	[-0.86, -0.60]
➖	file_to_blackhole	egress throughput	-0.95	[-3.44, +1.55]
➖	datadog_agent_remap_blackhole_acks	ingress throughput	-1.22	[-1.30, -1.13]
➖	datadog_agent_remap_datadog_logs	ingress throughput	-1.23	[-1.34, -1.12]
➖	syslog_regex_logs2metric_ddmetrics	ingress throughput	-1.41	[-1.50, -1.31]
➖	splunk_hec_route_s3	ingress throughput	-1.46	[-1.92, -1.00]
➖	syslog_humio_logs	ingress throughput	-2.62	[-2.74, -2.50]
➖	syslog_log2metric_splunk_hec_metrics	ingress throughput	-4.56	[-4.70, -4.41]
➖	syslog_loki	ingress throughput	-4.80	[-4.85, -4.75]
➖	syslog_splunk_hec_logs	ingress throughput	-4.96	[-5.06, -4.86]
❌	syslog_log2metric_tag_cardinality_limit_blackhole	ingress throughput	-6.06	[-6.18, -5.94]

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".

[github-actions]

Regression Detector Results

Run ID: 5681deea-26f8-4dfa-9ba5-bc7cdf745941
Baseline: 4a4fc2e
Comparison: eb67ee0
Total CPUs: 7

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

Significant changes in experiment optimization goals

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

perf	experiment	goal	Δ mean %	Δ mean % CI
❌	syslog_splunk_hec_logs	ingress throughput	-5.22	[-5.28, -5.16]
❌	syslog_log2metric_splunk_hec_metrics	ingress throughput	-5.43	[-5.59, -5.27]
❌	syslog_loki	ingress throughput	-7.18	[-7.25, -7.11]

Fine details of change detection per experiment

perf	experiment	goal	Δ mean %	Δ mean % CI
➖	otlp_http_to_blackhole	ingress throughput	+4.67	[+4.54, +4.81]
➖	splunk_hec_route_s3	ingress throughput	+2.70	[+2.22, +3.18]
➖	file_to_blackhole	egress throughput	+2.49	[+0.03, +4.96]
➖	otlp_grpc_to_blackhole	ingress throughput	+2.47	[+2.38, +2.56]
➖	http_text_to_http_json	ingress throughput	+2.44	[+2.32, +2.57]
➖	syslog_log2metric_humio_metrics	ingress throughput	+2.00	[+1.84, +2.17]
➖	socket_to_socket_blackhole	ingress throughput	+1.05	[+0.96, +1.13]
➖	http_to_http_acks	ingress throughput	+0.96	[-0.40, +2.33]
➖	datadog_agent_remap_datadog_logs	ingress throughput	+0.80	[+0.68, +0.92]
➖	fluent_elasticsearch	ingress throughput	+0.58	[+0.09, +1.06]
➖	http_to_s3	ingress throughput	+0.34	[+0.06, +0.62]
➖	datadog_agent_remap_blackhole_acks	ingress throughput	+0.10	[+0.00, +0.19]
➖	http_to_http_noack	ingress throughput	+0.09	[-0.01, +0.19]
➖	http_to_http_json	ingress throughput	+0.04	[-0.04, +0.12]
➖	splunk_hec_to_splunk_hec_logs_acks	ingress throughput	-0.00	[-0.14, +0.14]
➖	splunk_hec_indexer_ack_blackhole	ingress throughput	-0.01	[-0.15, +0.14]
➖	datadog_agent_remap_datadog_logs_acks	ingress throughput	-0.03	[-0.11, +0.06]
➖	enterprise_http_to_http	ingress throughput	-0.03	[-0.13, +0.06]
➖	splunk_hec_to_splunk_hec_logs_noack	ingress throughput	-0.05	[-0.16, +0.07]
➖	http_elasticsearch	ingress throughput	-0.15	[-0.22, -0.08]
➖	syslog_regex_logs2metric_ddmetrics	ingress throughput	-0.78	[-0.94, -0.62]
➖	datadog_agent_remap_blackhole	ingress throughput	-1.66	[-1.74, -1.57]
➖	syslog_humio_logs	ingress throughput	-1.68	[-1.80, -1.56]
➖	syslog_log2metric_tag_cardinality_limit_blackhole	ingress throughput	-2.83	[-2.96, -2.71]
❌	syslog_splunk_hec_logs	ingress throughput	-5.22	[-5.28, -5.16]
❌	syslog_log2metric_splunk_hec_metrics	ingress throughput	-5.43	[-5.59, -5.27]
❌	syslog_loki	ingress throughput	-7.18	[-7.25, -7.11]

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".

[github-actions]

Regression Detector Results

Run ID: b472ba00-ef5c-4d15-974d-c0c39c40e873
Baseline: 4a4fc2e
Comparison: b1d6043
Total CPUs: 7

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

Significant changes in experiment optimization goals

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

perf	experiment	goal	Δ mean %	Δ mean % CI
✅	otlp_http_to_blackhole	ingress throughput	+5.75	[+5.62, +5.88]
❌	syslog_log2metric_splunk_hec_metrics	ingress throughput	-5.07	[-5.21, -4.93]

Fine details of change detection per experiment

perf	experiment	goal	Δ mean %	Δ mean % CI
✅	otlp_http_to_blackhole	ingress throughput	+5.75	[+5.62, +5.88]
➖	syslog_log2metric_humio_metrics	ingress throughput	+1.86	[+1.78, +1.94]
➖	otlp_grpc_to_blackhole	ingress throughput	+1.33	[+1.24, +1.42]
➖	splunk_hec_route_s3	ingress throughput	+0.97	[+0.52, +1.42]
➖	http_text_to_http_json	ingress throughput	+0.56	[+0.45, +0.68]
➖	datadog_agent_remap_blackhole_acks	ingress throughput	+0.49	[+0.41, +0.58]
➖	socket_to_socket_blackhole	ingress throughput	+0.46	[+0.38, +0.54]
➖	datadog_agent_remap_datadog_logs_acks	ingress throughput	+0.23	[+0.15, +0.31]
➖	http_to_s3	ingress throughput	+0.15	[-0.13, +0.43]
➖	datadog_agent_remap_blackhole	ingress throughput	+0.14	[+0.04, +0.23]
➖	http_to_http_noack	ingress throughput	+0.13	[+0.05, +0.20]
➖	http_to_http_json	ingress throughput	+0.06	[-0.02, +0.14]
➖	splunk_hec_to_splunk_hec_logs_acks	ingress throughput	+0.00	[-0.14, +0.14]
➖	splunk_hec_indexer_ack_blackhole	ingress throughput	-0.01	[-0.15, +0.14]
➖	splunk_hec_to_splunk_hec_logs_noack	ingress throughput	-0.06	[-0.17, +0.06]
➖	http_to_http_acks	ingress throughput	-0.10	[-1.47, +1.26]
➖	enterprise_http_to_http	ingress throughput	-0.11	[-0.20, -0.01]
➖	datadog_agent_remap_datadog_logs	ingress throughput	-0.36	[-0.46, -0.25]
➖	fluent_elasticsearch	ingress throughput	-0.47	[-0.95, +0.01]
➖	file_to_blackhole	egress throughput	-0.49	[-2.99, +2.01]
➖	http_elasticsearch	ingress throughput	-0.96	[-1.03, -0.89]
➖	syslog_splunk_hec_logs	ingress throughput	-3.47	[-3.53, -3.42]
➖	syslog_regex_logs2metric_ddmetrics	ingress throughput	-3.88	[-3.97, -3.80]
➖	syslog_humio_logs	ingress throughput	-4.14	[-4.23, -4.04]
➖	syslog_log2metric_tag_cardinality_limit_blackhole	ingress throughput	-4.41	[-4.53, -4.29]
➖	syslog_loki	ingress throughput	-4.41	[-4.46, -4.36]
❌	syslog_log2metric_splunk_hec_metrics	ingress throughput	-5.07	[-5.21, -4.93]

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".

[github-actions]

Regression Detector Results

Run ID: 8bac803b-8bb3-46ab-b00a-d3ff6f4159a0
Baseline: 4a4fc2e
Comparison: 911824f
Total CPUs: 7

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

Significant changes in experiment optimization goals

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

perf	experiment	goal	Δ mean %	Δ mean % CI
✅	otlp_http_to_blackhole	ingress throughput	+5.21	[+5.08, +5.34]
❌	syslog_splunk_hec_logs	ingress throughput	-5.21	[-5.26, -5.16]
❌	syslog_log2metric_tag_cardinality_limit_blackhole	ingress throughput	-5.75	[-5.87, -5.63]

Fine details of change detection per experiment

perf	experiment	goal	Δ mean %	Δ mean % CI
✅	otlp_http_to_blackhole	ingress throughput	+5.21	[+5.08, +5.34]
➖	otlp_grpc_to_blackhole	ingress throughput	+1.86	[+1.77, +1.95]
➖	file_to_blackhole	egress throughput	+1.52	[-1.02, +4.06]
➖	syslog_log2metric_humio_metrics	ingress throughput	+1.35	[+1.25, +1.44]
➖	socket_to_socket_blackhole	ingress throughput	+1.22	[+1.15, +1.29]
➖	http_text_to_http_json	ingress throughput	+0.38	[+0.26, +0.51]
➖	syslog_regex_logs2metric_ddmetrics	ingress throughput	+0.30	[+0.17, +0.43]
➖	datadog_agent_remap_blackhole	ingress throughput	+0.29	[+0.20, +0.37]
➖	datadog_agent_remap_blackhole_acks	ingress throughput	+0.26	[+0.17, +0.35]
➖	http_to_http_noack	ingress throughput	+0.15	[+0.06, +0.24]
➖	datadog_agent_remap_datadog_logs_acks	ingress throughput	+0.12	[+0.04, +0.21]
➖	http_to_s3	ingress throughput	+0.08	[-0.20, +0.36]
➖	http_to_http_json	ingress throughput	+0.05	[-0.03, +0.13]
➖	splunk_hec_indexer_ack_blackhole	ingress throughput	+0.00	[-0.14, +0.15]
➖	splunk_hec_to_splunk_hec_logs_acks	ingress throughput	-0.00	[-0.14, +0.14]
➖	datadog_agent_remap_datadog_logs	ingress throughput	-0.00	[-0.11, +0.10]
➖	splunk_hec_to_splunk_hec_logs_noack	ingress throughput	-0.03	[-0.15, +0.08]
➖	enterprise_http_to_http	ingress throughput	-0.08	[-0.14, -0.03]
➖	http_to_http_acks	ingress throughput	-0.29	[-1.65, +1.07]
➖	splunk_hec_route_s3	ingress throughput	-0.62	[-1.07, -0.17]
➖	fluent_elasticsearch	ingress throughput	-0.79	[-1.27, -0.32]
➖	http_elasticsearch	ingress throughput	-0.85	[-0.92, -0.78]
➖	syslog_humio_logs	ingress throughput	-3.31	[-3.43, -3.20]
➖	syslog_log2metric_splunk_hec_metrics	ingress throughput	-3.85	[-3.98, -3.71]
➖	syslog_loki	ingress throughput	-4.87	[-4.92, -4.81]
❌	syslog_splunk_hec_logs	ingress throughput	-5.21	[-5.26, -5.16]
❌	syslog_log2metric_tag_cardinality_limit_blackhole	ingress throughput	-5.75	[-5.87, -5.63]

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".

[StephenWakely]

These regression tests should really not keep failing...

[github-actions]

Regression Detector Results

Run ID: 0d3c08af-975e-487d-b242-57a79b63824d
Baseline: 3da355b
Comparison: 7ce8fd5
Total CPUs: 7

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.

Fine details of change detection per experiment

perf	experiment	goal	Δ mean %	Δ mean % CI
➖	syslog_log2metric_splunk_hec_metrics	ingress throughput	+3.08	[+2.93, +3.22]
➖	syslog_splunk_hec_logs	ingress throughput	+1.50	[+1.45, +1.56]
➖	syslog_humio_logs	ingress throughput	+1.42	[+1.29, +1.55]
➖	datadog_agent_remap_datadog_logs_acks	ingress throughput	+1.36	[+1.28, +1.44]
➖	syslog_regex_logs2metric_ddmetrics	ingress throughput	+1.30	[+1.21, +1.38]
➖	syslog_log2metric_tag_cardinality_limit_blackhole	ingress throughput	+1.12	[+1.01, +1.23]
➖	fluent_elasticsearch	ingress throughput	+0.67	[+0.18, +1.15]
➖	syslog_loki	ingress throughput	+0.49	[+0.41, +0.57]
➖	datadog_agent_remap_blackhole_acks	ingress throughput	+0.41	[+0.30, +0.51]
➖	datadog_agent_remap_datadog_logs	ingress throughput	+0.31	[+0.20, +0.41]
➖	http_to_http_acks	ingress throughput	+0.28	[-1.08, +1.65]
➖	http_elasticsearch	ingress throughput	+0.25	[+0.18, +0.32]
➖	http_to_http_noack	ingress throughput	+0.12	[+0.02, +0.21]
➖	http_to_http_json	ingress throughput	+0.05	[-0.03, +0.13]
➖	splunk_hec_indexer_ack_blackhole	ingress throughput	+0.00	[-0.15, +0.15]
➖	splunk_hec_to_splunk_hec_logs_acks	ingress throughput	-0.00	[-0.14, +0.14]
➖	splunk_hec_to_splunk_hec_logs_noack	ingress throughput	-0.03	[-0.14, +0.08]
➖	enterprise_http_to_http	ingress throughput	-0.06	[-0.14, +0.02]
➖	http_to_s3	ingress throughput	-0.33	[-0.60, -0.05]
➖	otlp_grpc_to_blackhole	ingress throughput	-0.35	[-0.44, -0.26]
➖	syslog_log2metric_humio_metrics	ingress throughput	-0.89	[-0.98, -0.80]
➖	http_text_to_http_json	ingress throughput	-0.94	[-1.06, -0.83]
➖	file_to_blackhole	egress throughput	-1.01	[-3.46, +1.45]
➖	splunk_hec_route_s3	ingress throughput	-1.15	[-1.61, -0.69]
➖	socket_to_socket_blackhole	ingress throughput	-1.30	[-1.38, -1.21]
➖	datadog_agent_remap_blackhole	ingress throughput	-1.53	[-1.64, -1.42]
➖	otlp_http_to_blackhole	ingress throughput	-2.11	[-2.23, -1.99]

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".

[caibirdme]

It's weird... My code should only affect opentelemetry-source

[StephenWakely]

This is now failing with doc tests:

failures:

---- target/debug/build/opentelemetry-proto-d2ae8386e81e6113/out/opentelemetry.proto.trace.v1.rs - proto::trace::v1::Span::attributes (line 123) stdout ----
error: expected one of `.`, `;`, `?`, `}`, or an operator, found `:`
 --> target/debug/build/opentelemetry-proto-d2ae8386e81e6113/out/opentelemetry.proto.trace.v1.rs:124:19
  |
3 | "/http/user_agent": "Mozilla/5.0 (Macintosh; Intel Mac OS X 10_14_2) AppleWebKit/537.36 (KHTML, like Gecko) Chrome/71.0.3578.98 Safari/537.36"
  |                   ^ expected one of `.`, `;`, `?`, `}`, or an operator

error: aborting due to 1 previous error

On this doc comment in the generated proto file:

    /// attributes is a collection of key/value pairs. Note, global attributes
    /// like server name can be set using the resource API. Examples of attributes:
    ///
    ///      "/http/user_agent": "Mozilla/5.0 (Macintosh; Intel Mac OS X 10_14_2) AppleWebKit/537.36 (KHTML, like Gecko) Chrome/71.0.3578.98 Safari/537.36"
    ///      "/http/server_latency": 300
    ///      "example.com/my_attribute": true
    ///      "example.com/score": 10.239
    ///
    /// The OpenTelemetry API specification further restricts the allowed value types:
    /// <https://github.com/open-telemetry/opentelemetry-specification/blob/main/specification/common/README.md#attribute>
    /// Attribute keys MUST be unique (it is not allowed to have more than one
    /// attribute with the same key).
    #[prost(message, repeated, tag = "9")]

I'm not sure I understand why..

[StephenWakely]

The doc tests are failing because the comments here are indented by 4 spaces, so the doc tester thinks that it is Rust code in the comments. This PR didn't change that code, but I'm guessing that since nothing used it prior to this PR the code wasn't pulled in.

It's annoying to have to do, but I think the only way to fix this would be to update those lines in the comments so they only indent by 2 spaces.

Would you be able to make this change?

[github-actions]

Regression Detector Results

Run ID: 7db41c8d-c248-43c7-8f89-e4c83d33f959
Baseline: 3da355b
Comparison: d1d122e
Total CPUs: 7

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.

Fine details of change detection per experiment

perf	experiment	goal	Δ mean %	Δ mean % CI
➖	syslog_splunk_hec_logs	ingress throughput	+2.32	[+2.27, +2.37]
➖	syslog_loki	ingress throughput	+1.96	[+1.91, +2.01]
➖	syslog_humio_logs	ingress throughput	+1.88	[+1.76, +2.00]
➖	http_to_http_acks	ingress throughput	+1.24	[-0.13, +2.60]
➖	http_elasticsearch	ingress throughput	+1.21	[+1.13, +1.29]
➖	datadog_agent_remap_datadog_logs	ingress throughput	+0.85	[+0.75, +0.96]
➖	datadog_agent_remap_datadog_logs_acks	ingress throughput	+0.79	[+0.71, +0.87]
➖	fluent_elasticsearch	ingress throughput	+0.78	[+0.31, +1.26]
➖	syslog_regex_logs2metric_ddmetrics	ingress throughput	+0.71	[+0.63, +0.78]
➖	http_to_http_noack	ingress throughput	+0.15	[+0.06, +0.24]
➖	http_to_http_json	ingress throughput	+0.01	[-0.06, +0.08]
➖	splunk_hec_indexer_ack_blackhole	ingress throughput	-0.00	[-0.15, +0.14]
➖	splunk_hec_to_splunk_hec_logs_acks	ingress throughput	-0.00	[-0.14, +0.14]
➖	splunk_hec_to_splunk_hec_logs_noack	ingress throughput	-0.04	[-0.16, +0.07]
➖	enterprise_http_to_http	ingress throughput	-0.07	[-0.12, -0.02]
➖	syslog_log2metric_tag_cardinality_limit_blackhole	ingress throughput	-0.14	[-0.25, -0.02]
➖	http_to_s3	ingress throughput	-0.20	[-0.47, +0.08]
➖	syslog_log2metric_splunk_hec_metrics	ingress throughput	-0.37	[-0.50, -0.23]
➖	file_to_blackhole	egress throughput	-0.52	[-2.99, +1.95]
➖	syslog_log2metric_humio_metrics	ingress throughput	-0.68	[-0.79, -0.57]
➖	socket_to_socket_blackhole	ingress throughput	-0.71	[-0.79, -0.64]
➖	datadog_agent_remap_blackhole	ingress throughput	-0.90	[-1.00, -0.79]
➖	otlp_http_to_blackhole	ingress throughput	-0.95	[-1.09, -0.82]
➖	otlp_grpc_to_blackhole	ingress throughput	-0.98	[-1.07, -0.89]
➖	datadog_agent_remap_blackhole_acks	ingress throughput	-1.17	[-1.26, -1.07]
➖	http_text_to_http_json	ingress throughput	-1.27	[-1.40, -1.14]
➖	splunk_hec_route_s3	ingress throughput	-2.20	[-2.65, -1.74]

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".