feat(trace_exporter)!: add `restart_after_fork` configuration option [APMSP-3843] by VianneyRuhlmann · Pull Request #2263 · DataDog/libdatadog

VianneyRuhlmann · 2026-07-22T16:24:09Z

What does this PR do?

This adds a config option to the trace exporter to allow background workers to not be restarted in child. This used to be the default behavior and is now opt-in. This feature is only used in dd-trace-py so it will only be a breaking change there.

Note: This change also exposed that both the StatsExporter and OtlpStatsExporter were missing the reset implementation. It 's added as part of this PR.

Motivation

In python, the TraceExporter is currently recreated on forks, this is a legacy of the old thread architecture in python. To avoid restarting the workers in the child just to shut them down when recreating, we skip the restart on the trace exporter background workers. We want to change this behavior in dd-trace-py (DataDog/dd-trace-py#18929) as well as allow dd-trace-rb to use the workers in the intended way. Once dd-trace-py has transitioned this can be removed.

Additional Notes

Anything else we should know when reviewing?

How to test the change?

Describe here in detail how the change can be validated.

datadog-datadog-prod-us1-2 · 2026-07-22T16:32:25Z

Tests

🎉 All green!

🧪 All tests passed
❄️ No new flaky tests detected

🎯 Code Coverage (details)
• Patch Coverage: 60.00%
• Overall Coverage: 74.60% (-0.06%)

_{This comment will be updated automatically if new data arrives.

🔗 Commit SHA: 898bb80 | Docs | Datadog PR Page | Give us feedback!}

chatgpt-codex-connector

💡 Codex Review

Here are some automated review suggestions for this pull request.

Reviewed commit: 5f92c2c317

ℹ️ About Codex in GitHub

Codex has been enabled to automatically review pull requests in this repo. Reviews are triggered when you

Open a pull request for review
Mark a draft as ready
Comment "@codex review".

If Codex has suggestions, it will comment; otherwise it will react with 👍.

When you sign up for Codex through ChatGPT, Codex can also answer questions or update the PR, like "@codex address that feedback".

pr-commenter · 2026-07-22T16:36:58Z

Benchmarks

Comparison

Benchmark execution time: 2026-07-27 10:12:38

Comparing candidate commit 898bb80 in PR branch vianney/configure-restart-after-fork with baseline commit be0b543 in branch main.

Found 0 performance improvements and 0 performance regressions! Performance is the same for 24 metrics, 0 unstable metrics.

Explanation

This is an A/B test comparing a candidate commit's performance against that of a baseline commit. Performance changes are noted in the tables below as:

🟩 = significantly better candidate vs. baseline
🟥 = significantly worse candidate vs. baseline

We compute a confidence interval (CI) over the relative difference of means between metrics from the candidate and baseline commits, considering the baseline as the reference.

If the CI is entirely outside the configured SIGNIFICANT_IMPACT_THRESHOLD (or the deprecated UNCONFIDENCE_THRESHOLD), the change is considered significant.

Feel free to reach out to #apm-benchmarking-platform on Slack if you have any questions.

More details about the CI and significant changes

You can imagine this CI as a range of values that is likely to contain the true difference of means between the candidate and baseline commits.

CIs of the difference of means are often centered around 0%, because often changes are not that big:

---------------------------------(------|---^--------)-------------------------------->
                              -0.6%    0%  0.3%     +1.2%
                                 |          |        |
         lower bound of the CI --'          |        |
sample mean (center of the CI) -------------'        |
         upper bound of the CI ----------------------'

As described above, a change is considered significant if the CI is entirely outside the configured SIGNIFICANT_IMPACT_THRESHOLD (or the deprecated UNCONFIDENCE_THRESHOLD).

For instance, for an execution time metric, this confidence interval indicates a significantly worse performance:

----------------------------------------|---------|---(---------^---------)---------->
                                       0%        1%  1.3%      2.2%      3.1%
                                                  |   |         |         |
       significant impact threshold --------------'   |         |         |
                      lower bound of CI --------------'         |         |
       sample mean (center of the CI) --------------------------'         |
                      upper bound of CI ----------------------------------'

Benchmark execution time: 2026-07-27 10:09:19

Comparing candidate commit 898bb80 in PR branch vianney/configure-restart-after-fork with baseline commit be0b543 in branch main.

Found 0 performance improvements and 0 performance regressions! Performance is the same for 3 metrics, 0 unstable metrics.

Explanation

This is an A/B test comparing a candidate commit's performance against that of a baseline commit. Performance changes are noted in the tables below as:

🟩 = significantly better candidate vs. baseline
🟥 = significantly worse candidate vs. baseline

We compute a confidence interval (CI) over the relative difference of means between metrics from the candidate and baseline commits, considering the baseline as the reference.

If the CI is entirely outside the configured SIGNIFICANT_IMPACT_THRESHOLD (or the deprecated UNCONFIDENCE_THRESHOLD), the change is considered significant.

Feel free to reach out to #apm-benchmarking-platform on Slack if you have any questions.

More details about the CI and significant changes

You can imagine this CI as a range of values that is likely to contain the true difference of means between the candidate and baseline commits.

CIs of the difference of means are often centered around 0%, because often changes are not that big:

---------------------------------(------|---^--------)-------------------------------->
                              -0.6%    0%  0.3%     +1.2%
                                 |          |        |
         lower bound of the CI --'          |        |
sample mean (center of the CI) -------------'        |
         upper bound of the CI ----------------------'

As described above, a change is considered significant if the CI is entirely outside the configured SIGNIFICANT_IMPACT_THRESHOLD (or the deprecated UNCONFIDENCE_THRESHOLD).

For instance, for an execution time metric, this confidence interval indicates a significantly worse performance:

----------------------------------------|---------|---(---------^---------)---------->
                                       0%        1%  1.3%      2.2%      3.1%
                                                  |   |         |         |
       significant impact threshold --------------'   |         |         |
                      lower bound of CI --------------'         |         |
       sample mean (center of the CI) --------------------------'         |
                      upper bound of CI ----------------------------------'

Candidate

Candidate benchmark details

Group 1

cpu_model	git_commit_sha	git_commit_date	git_branch
Intel(R) Xeon(R) Platinum 8259CL CPU @ 2.50GHz	`898bb80`	1785145369	vianney/configure-restart-after-fork

scenario	metric	min	mean ± sd	median ± mad	p75	p95	p99	max	peak_to_median_ratio	skewness	kurtosis	cv	sem	runs	sample_size
trace_buffer/1_senders/10us_delay	execution_time	56.297ms	56.533ms ± 0.149ms	56.516ms ± 0.072ms	56.589ms	56.783ms	57.031ms	57.444ms	1.64%	2.029	7.855	0.26%	0.011ms	1	200
trace_buffer/1_senders/10us_delay	throughput	15667.548op/s	15919.905op/s ± 41.675op/s	15924.798op/s ± 20.415op/s	15944.256op/s	15970.948op/s	15977.293op/s	15986.647op/s	0.39%	-1.984	7.553	0.26%	2.947op/s	1	200
trace_buffer/1_senders/1us_delay	execution_time	50.026ms	50.082ms ± 0.028ms	50.081ms ± 0.011ms	50.090ms	50.135ms	50.185ms	50.192ms	0.22%	1.034	2.657	0.06%	0.002ms	1	200
trace_buffer/1_senders/1us_delay	throughput	17931.321op/s	17970.699op/s ± 9.966op/s	17970.941op/s ± 4.100op/s	17975.743op/s	17986.839op/s	17989.802op/s	17990.689op/s	0.11%	-1.028	2.640	0.06%	0.705op/s	1	200
trace_buffer/1_senders/no_delay	execution_time	330.805µs	339.973µs ± 3.279µs	340.219µs ± 2.105µs	341.919µs	343.934µs	349.695µs	355.959µs	4.63%	0.765	3.167	0.96%	0.232µs	1	200
trace_buffer/1_senders/no_delay	throughput	2528381.265op/s	2647514.174op/s ± 25361.292op/s	2645355.528op/s ± 16356.827op/s	2665152.725op/s	2684044.523op/s	2695879.106op/s	2720632.345op/s	2.85%	-0.638	2.694	0.96%	1793.314op/s	1	200
trace_buffer/2_senders/10us_delay	execution_time	56.697ms	57.102ms ± 0.159ms	57.094ms ± 0.100ms	57.187ms	57.388ms	57.513ms	57.571ms	0.84%	0.393	0.100	0.28%	0.011ms	1	200
trace_buffer/2_senders/10us_delay	throughput	31265.729op/s	31522.883op/s ± 87.518op/s	31526.835op/s ± 55.303op/s	31581.597op/s	31655.409op/s	31679.444op/s	31747.795op/s	0.70%	-0.377	0.084	0.28%	6.188op/s	1	200
trace_buffer/2_senders/1us_delay	execution_time	50.063ms	50.119ms ± 0.036ms	50.113ms ± 0.017ms	50.130ms	50.190ms	50.237ms	50.249ms	0.27%	1.244	1.494	0.07%	0.003ms	1	200
trace_buffer/2_senders/1us_delay	throughput	35821.730op/s	35914.377op/s ± 25.946op/s	35918.602op/s ± 12.466op/s	35932.781op/s	35944.870op/s	35948.633op/s	35954.393op/s	0.10%	-1.239	1.481	0.07%	1.835op/s	1	200
trace_buffer/2_senders/no_delay	execution_time	849.039µs	903.357µs ± 18.310µs	905.467µs ± 12.302µs	917.047µs	929.444µs	937.021µs	968.272µs	6.94%	-0.260	0.258	2.02%	1.295µs	1	200
trace_buffer/2_senders/no_delay	throughput	1858981.667op/s	1993387.953op/s ± 40670.822op/s	1987924.043op/s ± 26836.187op/s	2023226.010op/s	2064363.849op/s	2099316.075op/s	2120043.817op/s	6.65%	0.390	0.240	2.04%	2875.861op/s	1	200
trace_buffer/4_senders/10us_delay	execution_time	57.127ms	57.449ms ± 0.156ms	57.431ms ± 0.094ms	57.538ms	57.723ms	57.839ms	57.936ms	0.88%	0.498	0.249	0.27%	0.011ms	1	200
trace_buffer/4_senders/10us_delay	throughput	62137.317op/s	62664.467op/s ± 169.575op/s	62684.221op/s ± 102.684op/s	62782.481op/s	62911.449op/s	62989.204op/s	63017.130op/s	0.53%	-0.482	0.224	0.27%	11.991op/s	1	200
trace_buffer/4_senders/1us_delay	execution_time	50.114ms	50.170ms ± 0.038ms	50.162ms ± 0.016ms	50.180ms	50.252ms	50.284ms	50.364ms	0.40%	1.745	4.151	0.07%	0.003ms	1	200
trace_buffer/4_senders/1us_delay	throughput	71479.685op/s	71756.077op/s ± 53.809op/s	71767.456op/s ± 22.840op/s	71788.398op/s	71815.908op/s	71827.923op/s	71835.611op/s	0.09%	-1.738	4.113	0.07%	3.805op/s	1	200
trace_buffer/4_senders/no_delay	execution_time	2.235ms	2.446ms ± 0.058ms	2.449ms ± 0.042ms	2.488ms	2.536ms	2.553ms	2.599ms	6.12%	-0.351	0.016	2.38%	0.004ms	1	200
trace_buffer/4_senders/no_delay	throughput	1385398.896op/s	1472919.989op/s ± 35450.116op/s	1470120.363op/s ± 24732.385op/s	1496094.565op/s	1533863.662op/s	1552002.500op/s	1610490.171op/s	9.55%	0.491	0.276	2.40%	2506.702op/s	1	200
trace_buffer/8_senders/10us_delay	execution_time	57.655ms	57.980ms ± 0.216ms	57.908ms ± 0.112ms	58.151ms	58.398ms	58.487ms	58.539ms	1.09%	0.791	-0.463	0.37%	0.015ms	1	200
trace_buffer/8_senders/10us_delay	throughput	122994.330op/s	124181.796op/s ± 460.571op/s	124336.114op/s ± 241.217op/s	124539.396op/s	124705.422op/s	124814.231op/s	124879.998op/s	0.44%	-0.780	-0.481	0.37%	32.567op/s	1	200
trace_buffer/8_senders/1us_delay	execution_time	50.215ms	50.292ms ± 0.062ms	50.276ms ± 0.028ms	50.310ms	50.411ms	50.504ms	50.532ms	0.51%	1.695	3.070	0.12%	0.004ms	1	200
trace_buffer/8_senders/1us_delay	throughput	142485.121op/s	143162.895op/s ± 177.185op/s	143210.499op/s ± 79.973op/s	143273.139op/s	143345.823op/s	143371.840op/s	143383.037op/s	0.12%	-1.686	3.036	0.12%	12.529op/s	1	200
trace_buffer/8_senders/no_delay	execution_time	5.666ms	6.492ms ± 0.194ms	6.527ms ± 0.112ms	6.638ms	6.722ms	6.796ms	6.806ms	4.27%	-1.273	2.075	2.98%	0.014ms	1	200
trace_buffer/8_senders/no_delay	throughput	1057928.802op/s	1110001.663op/s ± 34537.748op/s	1103122.238op/s ± 18854.478op/s	1122697.384op/s	1175736.756op/s	1228475.886op/s	1270721.138op/s	15.19%	1.513	3.106	3.10%	2442.188op/s	1	200

scenario	metric	95% CI mean	Shapiro-Wilk pvalue	Ljung-Box pvalue (lag=1)	Dip test pvalue
trace_buffer/1_senders/10us_delay	execution_time	[56.513ms; 56.554ms] or [-0.036%; +0.036%]	None	None	None
trace_buffer/1_senders/10us_delay	throughput	[15914.130op/s; 15925.681op/s] or [-0.036%; +0.036%]	None	None	None
trace_buffer/1_senders/1us_delay	execution_time	[50.078ms; 50.085ms] or [-0.008%; +0.008%]	None	None	None
trace_buffer/1_senders/1us_delay	throughput	[17969.318op/s; 17972.080op/s] or [-0.008%; +0.008%]	None	None	None
trace_buffer/1_senders/no_delay	execution_time	[339.518µs; 340.427µs] or [-0.134%; +0.134%]	None	None	None
trace_buffer/1_senders/no_delay	throughput	[2643999.342op/s; 2651029.005op/s] or [-0.133%; +0.133%]	None	None	None
trace_buffer/2_senders/10us_delay	execution_time	[57.080ms; 57.124ms] or [-0.039%; +0.039%]	None	None	None
trace_buffer/2_senders/10us_delay	throughput	[31510.754op/s; 31535.012op/s] or [-0.038%; +0.038%]	None	None	None
trace_buffer/2_senders/1us_delay	execution_time	[50.114ms; 50.124ms] or [-0.010%; +0.010%]	None	None	None
trace_buffer/2_senders/1us_delay	throughput	[35910.781op/s; 35917.973op/s] or [-0.010%; +0.010%]	None	None	None
trace_buffer/2_senders/no_delay	execution_time	[900.819µs; 905.894µs] or [-0.281%; +0.281%]	None	None	None
trace_buffer/2_senders/no_delay	throughput	[1987751.369op/s; 1999024.538op/s] or [-0.283%; +0.283%]	None	None	None
trace_buffer/4_senders/10us_delay	execution_time	[57.428ms; 57.471ms] or [-0.038%; +0.038%]	None	None	None
trace_buffer/4_senders/10us_delay	throughput	[62640.965op/s; 62687.968op/s] or [-0.038%; +0.038%]	None	None	None
trace_buffer/4_senders/1us_delay	execution_time	[50.165ms; 50.175ms] or [-0.010%; +0.010%]	None	None	None
trace_buffer/4_senders/1us_delay	throughput	[71748.620op/s; 71763.534op/s] or [-0.010%; +0.010%]	None	None	None
trace_buffer/4_senders/no_delay	execution_time	[2.437ms; 2.454ms] or [-0.330%; +0.330%]	None	None	None
trace_buffer/4_senders/no_delay	throughput	[1468006.944op/s; 1477833.034op/s] or [-0.334%; +0.334%]	None	None	None
trace_buffer/8_senders/10us_delay	execution_time	[57.950ms; 58.010ms] or [-0.052%; +0.052%]	None	None	None
trace_buffer/8_senders/10us_delay	throughput	[124117.966op/s; 124245.627op/s] or [-0.051%; +0.051%]	None	None	None
trace_buffer/8_senders/1us_delay	execution_time	[50.284ms; 50.301ms] or [-0.017%; +0.017%]	None	None	None
trace_buffer/8_senders/1us_delay	throughput	[143138.339op/s; 143187.451op/s] or [-0.017%; +0.017%]	None	None	None
trace_buffer/8_senders/no_delay	execution_time	[6.466ms; 6.519ms] or [-0.413%; +0.413%]	None	None	None
trace_buffer/8_senders/no_delay	throughput	[1105215.064op/s; 1114788.263op/s] or [-0.431%; +0.431%]	None	None	None

Group 1

cpu_model	git_commit_sha	git_commit_date	git_branch
Intel(R) Xeon(R) Platinum 8259CL CPU @ 2.50GHz	`898bb80`	1785145369	vianney/configure-restart-after-fork

scenario	metric	min	mean ± sd	median ± mad	p75	p95	p99	max	peak_to_median_ratio	skewness	kurtosis	cv	sem	runs	sample_size
write only interface	execution_time	1.068µs	1.088µs ± 0.007µs	1.091µs ± 0.002µs	1.093µs	1.094µs	1.099µs	1.101µs	0.89%	-1.075	-0.077	0.67%	0.001µs	1	200

scenario	metric	95% CI mean	Shapiro-Wilk pvalue	Ljung-Box pvalue (lag=1)	Dip test pvalue
write only interface	execution_time	[1.087µs; 1.089µs] or [-0.094%; +0.094%]	None	None	None

Group 2

cpu_model	git_commit_sha	git_commit_date	git_branch
Intel(R) Xeon(R) Platinum 8259CL CPU @ 2.50GHz	`898bb80`	1785145369	vianney/configure-restart-after-fork

scenario	metric	min	mean ± sd	median ± mad	p75	p95	p99	max	peak_to_median_ratio	skewness	kurtosis	cv	sem	runs	sample_size
two way interface	execution_time	15.434µs	15.736µs ± 0.193µs	15.695µs ± 0.097µs	15.808µs	16.111µs	16.246µs	16.770µs	6.85%	1.519	3.984	1.23%	0.014µs	1	200

scenario	metric	95% CI mean	Shapiro-Wilk pvalue	Ljung-Box pvalue (lag=1)	Dip test pvalue
two way interface	execution_time	[15.709µs; 15.763µs] or [-0.170%; +0.170%]	None	None	None

Group 3

cpu_model	git_commit_sha	git_commit_date	git_branch
Intel(R) Xeon(R) Platinum 8259CL CPU @ 2.50GHz	`898bb80`	1785145369	vianney/configure-restart-after-fork

scenario	metric	min	mean ± sd	median ± mad	p75	p95	p99	max	peak_to_median_ratio	skewness	kurtosis	cv	sem	runs	sample_size
concentrator/add_spans_to_concentrator	execution_time	8.568ms	8.596ms ± 0.016ms	8.593ms ± 0.010ms	8.605ms	8.626ms	8.648ms	8.656ms	0.73%	1.045	1.257	0.19%	0.001ms	1	200

scenario	metric	95% CI mean	Shapiro-Wilk pvalue	Ljung-Box pvalue (lag=1)	Dip test pvalue
concentrator/add_spans_to_concentrator	execution_time	[8.594ms; 8.598ms] or [-0.026%; +0.026%]	None	None	None

Baseline

Baseline benchmark details

Group 1

cpu_model	git_commit_sha	git_commit_date	git_branch
Intel(R) Xeon(R) Platinum 8259CL CPU @ 2.50GHz	`be0b543`	1784918029	main

scenario	metric	min	mean ± sd	median ± mad	p75	p95	p99	max	peak_to_median_ratio	skewness	kurtosis	cv	sem	runs	sample_size
trace_buffer/1_senders/10us_delay	execution_time	56.288ms	56.509ms ± 0.120ms	56.485ms ± 0.067ms	56.559ms	56.710ms	56.872ms	57.270ms	1.39%	1.882	7.880	0.21%	0.009ms	1	200
trace_buffer/1_senders/10us_delay	throughput	15715.077op/s	15926.714op/s ± 33.769op/s	15933.555op/s ± 19.030op/s	15951.666op/s	15970.969op/s	15977.384op/s	15989.212op/s	0.35%	-1.842	7.591	0.21%	2.388op/s	1	200
trace_buffer/1_senders/1us_delay	execution_time	50.023ms	50.078ms ± 0.032ms	50.079ms ± 0.013ms	50.088ms	50.137ms	50.193ms	50.210ms	0.26%	0.949	2.133	0.06%	0.002ms	1	200
trace_buffer/1_senders/1us_delay	throughput	17924.739op/s	17971.929op/s ± 11.440op/s	17971.672op/s ± 4.721op/s	17978.127op/s	17988.753op/s	17990.069op/s	17991.745op/s	0.11%	-0.943	2.113	0.06%	0.809op/s	1	200
trace_buffer/1_senders/no_delay	execution_time	336.473µs	343.835µs ± 4.258µs	343.780µs ± 1.424µs	344.914µs	347.865µs	353.060µs	388.428µs	12.99%	6.096	59.818	1.24%	0.301µs	1	200
trace_buffer/1_senders/no_delay	throughput	2317031.100op/s	2617910.735op/s ± 30303.247op/s	2617954.546op/s ± 10887.948op/s	2630019.754op/s	2657757.966op/s	2662013.394op/s	2674807.048op/s	2.17%	-5.243	48.446	1.15%	2142.763op/s	1	200
trace_buffer/2_senders/10us_delay	execution_time	56.567ms	57.003ms ± 0.277ms	56.921ms ± 0.157ms	57.148ms	57.557ms	57.661ms	57.811ms	1.56%	0.856	-0.049	0.48%	0.020ms	1	200
trace_buffer/2_senders/10us_delay	throughput	31135.976op/s	31578.267op/s ± 152.700op/s	31622.769op/s ± 87.141op/s	31690.608op/s	31769.365op/s	31789.163op/s	31820.748op/s	0.63%	-0.838	-0.083	0.48%	10.798op/s	1	200
trace_buffer/2_senders/1us_delay	execution_time	50.056ms	50.112ms ± 0.034ms	50.104ms ± 0.018ms	50.130ms	50.181ms	50.204ms	50.217ms	0.23%	0.978	0.556	0.07%	0.002ms	1	200
trace_buffer/2_senders/1us_delay	throughput	35844.274op/s	35919.715op/s ± 24.179op/s	35925.538op/s ± 13.086op/s	35936.826op/s	35950.327op/s	35957.858op/s	35959.433op/s	0.09%	-0.974	0.549	0.07%	1.710op/s	1	200
trace_buffer/2_senders/no_delay	execution_time	848.601µs	905.428µs ± 16.925µs	906.684µs ± 9.925µs	916.207µs	931.664µs	942.521µs	947.978µs	4.55%	-0.392	0.566	1.86%	1.197µs	1	200
trace_buffer/2_senders/no_delay	throughput	1898778.353op/s	1988706.631op/s ± 37495.912op/s	1985256.668op/s ± 21727.582op/s	2007924.905op/s	2055799.742op/s	2087188.843op/s	2121137.647op/s	6.84%	0.529	0.742	1.88%	2651.361op/s	1	200
trace_buffer/4_senders/10us_delay	execution_time	57.101ms	57.398ms ± 0.158ms	57.376ms ± 0.083ms	57.449ms	57.760ms	57.855ms	57.958ms	1.01%	1.026	1.196	0.27%	0.011ms	1	200
trace_buffer/4_senders/10us_delay	throughput	62114.045op/s	62720.495op/s ± 172.200op/s	62744.015op/s ± 90.825op/s	62841.754op/s	62957.374op/s	62984.990op/s	63046.465op/s	0.48%	-1.008	1.153	0.27%	12.176op/s	1	200
trace_buffer/4_senders/1us_delay	execution_time	50.120ms	50.167ms ± 0.032ms	50.158ms ± 0.013ms	50.178ms	50.232ms	50.279ms	50.319ms	0.32%	1.703	3.648	0.06%	0.002ms	1	200
trace_buffer/4_senders/1us_delay	throughput	71543.918op/s	71760.564op/s ± 46.275op/s	71773.621op/s ± 19.138op/s	71786.615op/s	71813.206op/s	71826.323op/s	71827.943op/s	0.08%	-1.697	3.622	0.06%	3.272op/s	1	200
trace_buffer/4_senders/no_delay	execution_time	2.094ms	2.365ms ± 0.092ms	2.372ms ± 0.055ms	2.425ms	2.505ms	2.542ms	2.552ms	7.59%	-0.366	-0.110	3.88%	0.007ms	1	200
trace_buffer/4_senders/no_delay	throughput	1410821.461op/s	1524333.630op/s ± 60461.431op/s	1517870.486op/s ± 35462.407op/s	1554392.951op/s	1639852.489op/s	1685802.375op/s	1718977.397op/s	13.25%	0.578	0.192	3.96%	4275.269op/s	1	200
trace_buffer/8_senders/10us_delay	execution_time	57.639ms	57.947ms ± 0.239ms	57.851ms ± 0.096ms	58.082ms	58.405ms	58.590ms	58.965ms	1.93%	1.282	1.254	0.41%	0.017ms	1	200
trace_buffer/8_senders/10us_delay	throughput	122105.709op/s	124252.534op/s ± 509.631op/s	124457.415op/s ± 207.587op/s	124606.115op/s	124777.348op/s	124876.365op/s	124916.495op/s	0.37%	-1.262	1.165	0.41%	36.036op/s	1	200
trace_buffer/8_senders/1us_delay	execution_time	50.212ms	50.297ms ± 0.057ms	50.279ms ± 0.033ms	50.328ms	50.403ms	50.444ms	50.540ms	0.52%	1.326	2.465	0.11%	0.004ms	1	200
trace_buffer/8_senders/1us_delay	throughput	142460.447op/s	143149.123op/s ± 160.876op/s	143200.981op/s ± 93.933op/s	143257.416op/s	143335.774op/s	143378.975op/s	143391.372op/s	0.13%	-1.316	2.424	0.11%	11.376op/s	1	200
trace_buffer/8_senders/no_delay	execution_time	5.964ms	6.535ms ± 0.126ms	6.554ms ± 0.070ms	6.616ms	6.686ms	6.731ms	6.783ms	3.49%	-1.512	3.890	1.92%	0.009ms	1	200
trace_buffer/8_senders/no_delay	throughput	1061486.403op/s	1102253.869op/s ± 21941.135op/s	1098504.864op/s ± 11712.560op/s	1111169.116op/s	1137533.224op/s	1178598.029op/s	1207228.482op/s	9.90%	1.732	4.934	1.99%	1551.473op/s	1	200

scenario	metric	95% CI mean	Shapiro-Wilk pvalue	Ljung-Box pvalue (lag=1)	Dip test pvalue
trace_buffer/1_senders/10us_delay	execution_time	[56.492ms; 56.526ms] or [-0.030%; +0.030%]	None	None	None
trace_buffer/1_senders/10us_delay	throughput	[15922.033op/s; 15931.394op/s] or [-0.029%; +0.029%]	None	None	None
trace_buffer/1_senders/1us_delay	execution_time	[50.074ms; 50.083ms] or [-0.009%; +0.009%]	None	None	None
trace_buffer/1_senders/1us_delay	throughput	[17970.343op/s; 17973.514op/s] or [-0.009%; +0.009%]	None	None	None
trace_buffer/1_senders/no_delay	execution_time	[343.244µs; 344.425µs] or [-0.172%; +0.172%]	None	None	None
trace_buffer/1_senders/no_delay	throughput	[2613710.997op/s; 2622110.474op/s] or [-0.160%; +0.160%]	None	None	None
trace_buffer/2_senders/10us_delay	execution_time	[56.964ms; 57.041ms] or [-0.067%; +0.067%]	None	None	None
trace_buffer/2_senders/10us_delay	throughput	[31557.105op/s; 31599.430op/s] or [-0.067%; +0.067%]	None	None	None
trace_buffer/2_senders/1us_delay	execution_time	[50.107ms; 50.116ms] or [-0.009%; +0.009%]	None	None	None
trace_buffer/2_senders/1us_delay	throughput	[35916.364op/s; 35923.066op/s] or [-0.009%; +0.009%]	None	None	None
trace_buffer/2_senders/no_delay	execution_time	[903.083µs; 907.774µs] or [-0.259%; +0.259%]	None	None	None
trace_buffer/2_senders/no_delay	throughput	[1983510.058op/s; 1993903.204op/s] or [-0.261%; +0.261%]	None	None	None
trace_buffer/4_senders/10us_delay	execution_time	[57.376ms; 57.420ms] or [-0.038%; +0.038%]	None	None	None
trace_buffer/4_senders/10us_delay	throughput	[62696.630op/s; 62744.361op/s] or [-0.038%; +0.038%]	None	None	None
trace_buffer/4_senders/1us_delay	execution_time	[50.162ms; 50.171ms] or [-0.009%; +0.009%]	None	None	None
trace_buffer/4_senders/1us_delay	throughput	[71754.151op/s; 71766.977op/s] or [-0.009%; +0.009%]	None	None	None
trace_buffer/4_senders/no_delay	execution_time	[2.353ms; 2.378ms] or [-0.540%; +0.540%]	None	None	None
trace_buffer/4_senders/no_delay	throughput	[1515954.257op/s; 1532713.003op/s] or [-0.550%; +0.550%]	None	None	None
trace_buffer/8_senders/10us_delay	execution_time	[57.914ms; 57.981ms] or [-0.057%; +0.057%]	None	None	None
trace_buffer/8_senders/10us_delay	throughput	[124181.904op/s; 124323.164op/s] or [-0.057%; +0.057%]	None	None	None
trace_buffer/8_senders/1us_delay	execution_time	[50.289ms; 50.305ms] or [-0.016%; +0.016%]	None	None	None
trace_buffer/8_senders/1us_delay	throughput	[143126.828op/s; 143171.419op/s] or [-0.016%; +0.016%]	None	None	None
trace_buffer/8_senders/no_delay	execution_time	[6.517ms; 6.552ms] or [-0.267%; +0.267%]	None	None	None
trace_buffer/8_senders/no_delay	throughput	[1099213.039op/s; 1105294.700op/s] or [-0.276%; +0.276%]	None	None	None

Group 1

cpu_model	git_commit_sha	git_commit_date	git_branch
Intel(R) Xeon(R) Platinum 8259CL CPU @ 2.50GHz	`be0b543`	1784918029	main

scenario	metric	min	mean ± sd	median ± mad	p75	p95	p99	max	peak_to_median_ratio	skewness	kurtosis	cv	sem	runs	sample_size
write only interface	execution_time	1.060µs	1.075µs ± 0.008µs	1.074µs ± 0.005µs	1.080µs	1.090µs	1.095µs	1.101µs	2.56%	0.624	0.295	0.74%	0.001µs	1	200

scenario	metric	95% CI mean	Shapiro-Wilk pvalue	Ljung-Box pvalue (lag=1)	Dip test pvalue
write only interface	execution_time	[1.074µs; 1.076µs] or [-0.102%; +0.102%]	None	None	None

Group 2

cpu_model	git_commit_sha	git_commit_date	git_branch
Intel(R) Xeon(R) Platinum 8259CL CPU @ 2.50GHz	`be0b543`	1784918029	main

scenario	metric	min	mean ± sd	median ± mad	p75	p95	p99	max	peak_to_median_ratio	skewness	kurtosis	cv	sem	runs	sample_size
two way interface	execution_time	15.349µs	15.706µs ± 0.168µs	15.674µs ± 0.102µs	15.790µs	16.032µs	16.127µs	16.344µs	4.28%	0.720	0.573	1.07%	0.012µs	1	200

scenario	metric	95% CI mean	Shapiro-Wilk pvalue	Ljung-Box pvalue (lag=1)	Dip test pvalue
two way interface	execution_time	[15.682µs; 15.729µs] or [-0.148%; +0.148%]	None	None	None

Group 3

cpu_model	git_commit_sha	git_commit_date	git_branch
Intel(R) Xeon(R) Platinum 8259CL CPU @ 2.50GHz	`be0b543`	1784918029	main

scenario	metric	min	mean ± sd	median ± mad	p75	p95	p99	max	peak_to_median_ratio	skewness	kurtosis	cv	sem	runs	sample_size
concentrator/add_spans_to_concentrator	execution_time	8.674ms	8.706ms ± 0.018ms	8.704ms ± 0.013ms	8.718ms	8.736ms	8.764ms	8.766ms	0.71%	0.749	0.486	0.21%	0.001ms	1	200

scenario	metric	95% CI mean	Shapiro-Wilk pvalue	Ljung-Box pvalue (lag=1)	Dip test pvalue
concentrator/add_spans_to_concentrator	execution_time	[8.703ms; 8.708ms] or [-0.029%; +0.029%]	None	None	None

dd-octo-sts · 2026-07-22T17:04:55Z

Artifact Size Benchmark Report

aarch64-alpine-linux-musl

Artifact	Baseline	Commit	Change
/aarch64-alpine-linux-musl/lib/libdatadog_profiling.a	88.18 MB	88.18 MB	+0% (+6.13 KB) 👌
/aarch64-alpine-linux-musl/lib/libdatadog_profiling.so	8.13 MB	8.13 MB	0% (0 B) 👌

aarch64-unknown-linux-gnu

Artifact	Baseline	Commit	Change
/aarch64-unknown-linux-gnu/lib/libdatadog_profiling.so	10.93 MB	10.93 MB	+0% (+992 B) 👌
/aarch64-unknown-linux-gnu/lib/libdatadog_profiling.a	99.46 MB	99.46 MB	+0% (+2.46 KB) 👌

libdatadog-x64-windows

Artifact	Baseline	Commit	Change
/libdatadog-x64-windows/debug/dynamic/datadog_profiling_ffi.dll	26.59 MB	26.59 MB	+0% (+1.00 KB) 👌
/libdatadog-x64-windows/debug/dynamic/datadog_profiling_ffi.lib	89.18 KB	89.18 KB	0% (0 B) 👌
/libdatadog-x64-windows/debug/dynamic/datadog_profiling_ffi.pdb	191.30 MB	191.29 MB	-0% (-8.00 KB) 👌
/libdatadog-x64-windows/debug/static/datadog_profiling_ffi.lib	1.04 GB	1.04 GB	+0% (+9.87 KB) 👌
/libdatadog-x64-windows/release/dynamic/datadog_profiling_ffi.dll	8.66 MB	8.66 MB	+.01% (+1.00 KB) 🔍
/libdatadog-x64-windows/release/dynamic/datadog_profiling_ffi.lib	89.18 KB	89.18 KB	0% (0 B) 👌
/libdatadog-x64-windows/release/dynamic/datadog_profiling_ffi.pdb	25.36 MB	25.36 MB	0% (0 B) 👌
/libdatadog-x64-windows/release/static/datadog_profiling_ffi.lib	50.55 MB	50.56 MB	+.01% (+8.69 KB) 🔍

libdatadog-x86-windows

Artifact	Baseline	Commit	Change
/libdatadog-x86-windows/debug/dynamic/datadog_profiling_ffi.dll	23.16 MB	23.16 MB	+0% (+1.50 KB) 👌
/libdatadog-x86-windows/debug/dynamic/datadog_profiling_ffi.lib	90.58 KB	90.58 KB	0% (0 B) 👌
/libdatadog-x86-windows/debug/dynamic/datadog_profiling_ffi.pdb	195.94 MB	195.94 MB	0% (0 B) 👌
/libdatadog-x86-windows/debug/static/datadog_profiling_ffi.lib	1.03 GB	1.03 GB	+0% (+10.27 KB) 👌
/libdatadog-x86-windows/release/dynamic/datadog_profiling_ffi.dll	6.68 MB	6.68 MB	+.02% (+1.50 KB) 🔍
/libdatadog-x86-windows/release/dynamic/datadog_profiling_ffi.lib	90.58 KB	90.58 KB	0% (0 B) 👌
/libdatadog-x86-windows/release/dynamic/datadog_profiling_ffi.pdb	27.26 MB	27.26 MB	0% (0 B) 👌
/libdatadog-x86-windows/release/static/datadog_profiling_ffi.lib	48.13 MB	48.14 MB	+.01% (+9.13 KB) 🔍

x86_64-alpine-linux-musl

Artifact	Baseline	Commit	Change
/x86_64-alpine-linux-musl/lib/libdatadog_profiling.a	78.64 MB	78.64 MB	+0% (+4.59 KB) 👌
/x86_64-alpine-linux-musl/lib/libdatadog_profiling.so	9.05 MB	9.06 MB	+.04% (+4.00 KB) 🔍

x86_64-unknown-linux-gnu

Artifact	Baseline	Commit	Change
/x86_64-unknown-linux-gnu/lib/libdatadog_profiling.a	94.14 MB	94.14 MB	+0% (+3.02 KB) 👌
/x86_64-unknown-linux-gnu/lib/libdatadog_profiling.so	11.02 MB	11.02 MB	+.04% (+4.73 KB) 🔍

github-actions · 2026-07-23T12:54:22Z

Clippy Allow Annotation Report

Tracked Clippy allow annotations changed vs main: ✅ -2 (4 → 2)

Rule	Base	PR	Δ
unwrap_used	4	2	✅ -2

By file and crate

By file

File	Base	PR	Δ
`libdd-data-pipeline/src/otlp/metrics.rs`	1	0	✅ -1
`libdd-trace-stats/src/stats_exporter.rs`	1	0	✅ -1

By crate

Crate	Base	PR	Δ
`libdd-data-pipeline`	5	4	✅ -1
`libdd-trace-stats`	1	0	✅ -1

About This Report

This report tracks Clippy allow annotations for specific rules, showing how they've changed in this PR. Decreasing the number of these annotations generally improves code quality. Panic-inducing macros in particular should be avoided. In the future, this report may become a PR-blocking quality gate.

yannham · 2026-07-23T16:37:26Z

+        let _ = self
+            .concentrator
+            .lock_or_panic()
+            .flush(SystemTime::now(), true);


Should the child flush the concentrator though? I would expect to leave the original concentrator entirely to the parent, and just re-initialize it here to something new.

This effectively resets the concentrator state by force flushing, the buckets are dropped not sent. Re-initializing the concentrator from the stats exporter would require to store all the parameters required to create the concentrator. Since the concentrator config can be updated by the TraceExporter this also means we have to update these parameters and update the concentrator handle in the TraceExporter.

yannham · 2026-07-23T16:39:03Z

Oh, and I forget: how do we make sure we clean this once dd-trace-py doesn't need it anymore? Should we have KTLO ticket or sth?

gh-worker-devflow-routing-ef8351 · 2026-07-27T14:20:27Z

View all feedbacks in Devflow UI.

2026-07-27 14:20:26 UTC ℹ️ Start processing command devflow:merge

2026-07-27 14:20:31 UTC ℹ️ MergeQueue: pull request added to the queue

The expected merge time in main is approximately 1h (p90).

2026-07-27 15:31:25 UTC ❌ MergeQueue: The checks failed on this merge request

Tests failed on this commit 27b6f10:

allchecks

What to do next?

Investigate the failures and when ready, re-add your pull request to the queue!
If your PR checks are green, try to rebase/merge. It might be because the CI run is a bit old.
Any question, go check the FAQ.

feat(trace_exporter)!: expose restart_after_fork option

5f92c2c

github-actions Bot added the data-pipeline label Jul 22, 2026

VianneyRuhlmann marked this pull request as ready for review July 22, 2026 16:25

VianneyRuhlmann requested a review from a team as a code owner July 22, 2026 16:25

chatgpt-codex-connector Bot reviewed Jul 22, 2026

View reviewed changes

Comment thread libdd-data-pipeline/src/trace_exporter/builder.rs

Comment thread libdd-data-pipeline/src/trace_exporter/builder.rs Outdated

Comment thread libdd-data-pipeline/src/trace_exporter/stats.rs

feat(stats): implement reset on stats exporter

2b4d8de

VianneyRuhlmann requested review from a team as code owners July 23, 2026 12:51

VianneyRuhlmann requested review from mabdinur and removed request for a team July 23, 2026 12:51

update missed spawn_worker

ca5e3de

VianneyRuhlmann changed the title ~~feat!: expose restart_after_fork option~~ feat(trace_exporter)!: add restart_after_fork configuration option [APMSP-3843] Jul 23, 2026

yannham approved these changes Jul 23, 2026

View reviewed changes

doc(trace_exporter): add todo with jira ticket

898bb80

VianneyRuhlmann added the devflow:merge label Jul 27, 2026 — with Graphite App

gh-worker-dd-devflow-36fce6 Bot added mergequeue-status: queued mergequeue-status: in_progress mergequeue-status: rejected and removed mergequeue-status: queued mergequeue-status: in_progress labels Jul 27, 2026

gh-worker-devflow-routing-ef8351 Bot removed the devflow:merge label Jul 27, 2026

gh-worker-dd-devflow-36fce6 Bot added mergequeue-status: queued and removed mergequeue-status: rejected mergequeue-status: queued labels Jul 27, 2026

gh-worker-dd-devflow-36fce6 Bot added the mergequeue-status: in_progress label Jul 27, 2026

gh-worker-dd-mergequeue-cf854d Bot merged commit 3f4be28 into main Jul 27, 2026
175 checks passed

gh-worker-dd-devflow-36fce6 Bot removed the mergequeue-status: in_progress label Jul 27, 2026

gh-worker-dd-mergequeue-cf854d Bot deleted the vianney/configure-restart-after-fork branch July 27, 2026 16:25

gh-worker-dd-devflow-36fce6 Bot added the mergequeue-status: done label Jul 27, 2026

Uh oh!

Conversation

VianneyRuhlmann commented Jul 22, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

What does this PR do?

Motivation

Additional Notes

How to test the change?

Uh oh!

datadog-datadog-prod-us1-2 Bot commented Jul 22, 2026 • edited by datadog-prod-us1-4 Bot Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

chatgpt-codex-connector Bot left a comment

Choose a reason for hiding this comment

💡 Codex Review

Uh oh!

Uh oh!

Uh oh!

Uh oh!

pr-commenter Bot commented Jul 22, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Benchmarks

Comparison

Explanation

More details about the CI and significant changes

Explanation

More details about the CI and significant changes

Candidate

Group 1

Group 1

Group 2

Group 3

Baseline

Group 1

Group 1

Group 2

Group 3

Uh oh!

dd-octo-sts Bot commented Jul 22, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Artifact Size Benchmark Report

Uh oh!

github-actions Bot commented Jul 23, 2026

Clippy Allow Annotation Report

About This Report

Uh oh!

yannham Jul 23, 2026

Choose a reason for hiding this comment

Uh oh!

VianneyRuhlmann Jul 27, 2026

Choose a reason for hiding this comment

Uh oh!

yannham commented Jul 23, 2026

Uh oh!

gh-worker-devflow-routing-ef8351 Bot commented Jul 27, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

What to do next?

Uh oh!

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

2 participants

VianneyRuhlmann commented Jul 22, 2026 •

edited

Loading

datadog-datadog-prod-us1-2 Bot commented Jul 22, 2026 •

edited by datadog-prod-us1-4 Bot

Loading

pr-commenter Bot commented Jul 22, 2026 •

edited

Loading

dd-octo-sts Bot commented Jul 22, 2026 •

edited

Loading

gh-worker-devflow-routing-ef8351 Bot commented Jul 27, 2026 •

edited

Loading