[poc] disable adaptive sampling on observer pipeline sources#52054
[poc] disable adaptive sampling on observer pipeline sources#52054CelianR wants to merge 2 commits into
Conversation
When logs_config.experimental_adaptive_sampling.enabled is true globally, the decoder falls back to that flag for any source with ExperimentalAdaptiveSampling == nil. logssource sources never set this field, so they were silently inheriting the AdaptiveSampler and dropping logs before the anomaly detection observer could see them. Stamp Enabled=false on every source that enters the logssource pipeline: - source_provider.handleSet for generic workloadmeta container sources - adSourceManager.AddSource for all AD-scheduled sources The per-source pointer takes precedence over the global flag in resolveAdaptiveSamplerEnabled, so the decoder always picks NoopSampler for logssource sources regardless of the global setting. The main agent pipeline is unaffected. Co-authored-by: Cursor <cursoragent@cursor.com>
CelianR
added
changelog/no-changelog
|
@codex review |
CelianR
changed the title
![chatgpt-codex-connector[bot]](https://avatars.githubusercontent.com/in/1144995?s=60&v=4){kind=small}
There was a problem hiding this comment.
💡 Codex Review
Here are some automated review suggestions for this pull request.
Reviewed commit: cb95343309
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".
| func disableAdaptiveSampling(cfg *logsconfig.LogsConfig) { | ||
| disabled := false | ||
| if cfg.ExperimentalAdaptiveSampling == nil { | ||
| cfg.ExperimentalAdaptiveSampling = &logsconfig.SourceAdaptiveSamplingOptions{} | ||
| } | ||
| cfg.ExperimentalAdaptiveSampling.Enabled = &disabled |
There was a problem hiding this comment.
When anomaly_detection.logs.kubelet.enabled is on, registerKubeletJournaldSource adds the kubelet journald LogSource directly to logSources without calling this helper, so with global logs_config.experimental_adaptive_sampling.enabled the kubelet.service logs are still sampled before they reach the observer. This leaves one observer logssource path with the old behavior despite the new unsampled-stream requirement; apply the same override when constructing the kubelet source too.
Useful? React with 👍 / 👎.
|
Tests that when global adaptive sampling is enabled, a source with ExperimentalAdaptiveSampling.Enabled=false (as stamped by disableAdaptiveSampling on all logssource sources) passes every message through, while a source with no override is rate-limited by the AdaptiveSampler. Two sub-tests: - sampling enabled: logssource decoder receives all N messages, main agent decoder drops all but the first burst credit - sampling disabled: both decoders receive all N messages Co-authored-by: Cursor <cursoragent@cursor.com>
|
@codex review |
There was a problem hiding this comment.
💡 Codex Review
Here are some automated review suggestions for this pull request.
Reviewed commit: edfcfddee7
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".
| // global logs_config.experimental_adaptive_sampling.enabled flag. The observer | ||
| // pipeline must receive an unsampled stream; dropping logs here would cause the | ||
| // anomaly detection engine to miss anomalies hidden in suppressed patterns. | ||
| func disableAdaptiveSampling(cfg *logsconfig.LogsConfig) { |
There was a problem hiding this comment.
In kubelet+systemd builds, registerKubeletJournaldSource creates and adds the kubelet journald source directly (comp/anomalydetection/logssource/impl/kubelet_source.go:20-26) and never calls this helper or sets ExperimentalAdaptiveSampling. When logs_config.experimental_adaptive_sampling.enabled is true, the decoder falls back to that global flag for a nil source option, so the observer pipeline still samples kubelet logs even though this override is intended to make logssource streams unsampled. Please apply the same override when creating the kubelet source as well.
Useful? React with 👍 / 👎.
Files inventory check summaryFile checks results against ancestor bcfd3ddc: Results for datadog-agent_7.81.0~devel.git.581.edfcfdd.pipeline.117902459-1_amd64.deb:No change detected |
Static quality checks✅ Please find below the results from static quality gates Successful checksInfo
22 successful checks with minimal change (< 2 KiB)
|
Regression DetectorRegression Detector ResultsMetrics dashboard Baseline: bcfd3dd Optimization Goals: ✅ No significant changes detected
|
| perf | experiment | goal | Δ mean % | Δ mean % CI | trials | links |
|---|---|---|---|---|---|---|
| ➖ | quality_gate_metrics_logs | memory utilization | +1.23 | [+0.97, +1.48] | 1 | Logs bounds checks dashboard |
| ➖ | quality_gate_idle | memory utilization | +0.34 | [+0.29, +0.39] | 1 | Logs bounds checks dashboard |
| ➖ | quality_gate_idle_all_features | memory utilization | -0.16 | [-0.20, -0.13] | 1 | Logs bounds checks dashboard |
| ➖ | quality_gate_logs | % cpu utilization | -0.29 | [-1.33, +0.76] | 1 | Logs bounds checks dashboard |
Bounds Checks: ✅ Passed
| perf | experiment | bounds_check_name | replicates_passed | observed_value | links |
|---|---|---|---|---|---|
| ✅ | quality_gate_idle | intake_connections | 10/10 | 3 ≤ 4 | bounds checks dashboard |
| ✅ | quality_gate_idle | memory_usage | 10/10 | 144.93MiB ≤ 147MiB | bounds checks dashboard |
| ✅ | quality_gate_idle | total_bytes_received | 10/10 | 733.47KiB ≤ 819.20KiB | bounds checks dashboard |
| ✅ | quality_gate_idle_all_features | intake_connections | 10/10 | 3 ≤ 4 | bounds checks dashboard |
| ✅ | quality_gate_idle_all_features | memory_usage | 10/10 | 473.92MiB ≤ 495MiB | bounds checks dashboard |
| ✅ | quality_gate_idle_all_features | total_bytes_received | 10/10 | 1.12MiB ≤ 1.25MiB | bounds checks dashboard |
| ✅ | quality_gate_logs | intake_connections | 10/10 | 4 ≤ 6 | bounds checks dashboard |
| ✅ | quality_gate_logs | memory_usage | 10/10 | 180.78MiB ≤ 195MiB | bounds checks dashboard |
| ✅ | quality_gate_logs | missed_bytes | 10/10 | 0B = 0B | bounds checks dashboard |
| ✅ | quality_gate_logs | total_bytes_received | 10/10 | 264.39MiB ≤ 292MiB | bounds checks dashboard |
| ✅ | quality_gate_metrics_logs | cpu_usage | 10/10 | 345.38 ≤ 2000 | bounds checks dashboard |
| ✅ | quality_gate_metrics_logs | intake_connections | 10/10 | 3 ≤ 6 | bounds checks dashboard |
| ✅ | quality_gate_metrics_logs | memory_usage | 10/10 | 395.03MiB ≤ 430MiB | bounds checks dashboard |
| ✅ | quality_gate_metrics_logs | missed_bytes | 10/10 | 0B = 0B | bounds checks dashboard |
| ✅ | quality_gate_metrics_logs | total_bytes_received | 10/10 | 0.94GiB ≤ 1.04GiB | bounds checks dashboard |
Explanation
Confidence level: 90.00%
Effect size tolerance: |Δ mean %| ≥ 5.00%
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
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:
-
Its estimated |Δ mean %| ≥ 5.00%, indicating the change is big enough to merit a closer look.
-
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.
-
Its configuration does not mark it "erratic".
Replicate Execution Details
We run multiple replicates for each experiment/variant. However, we allow replicates to be automatically retried if there are any failures, up to 8 times, at which point the replicate is marked dead and we are unable to run analysis for the entire experiment. We call each of these attempts at running replicates a replicate execution. This section lists all replicate executions that failed due to the target crashing or being oom killed.
Note: In the below tables we bucket failures by experiment, variant, and failure type. For each of these buckets we list out the replicate indexes that failed with an annotation signifying how many times said replicate failed with the given failure mode. In the below example the baseline variant of the experiment named experiment_with_failures had two replicates that failed by oom kills. Replicate 0, which failed 8 executions, and replicate 1 which failed 6 executions, all with the same failure mode.
| Experiment | Variant | Replicates | Failure | Logs | Debug Dashboard |
|---|---|---|---|---|---|
| experiment_with_failures | baseline | 0 (x8) 1 (x6) | Oom killed | Debug Dashboard |
The debug dashboard links will take you to a debugging dashboard specifically designed to investigate replicate execution failures.
❌ Retried Profiling Replicate Execution Failures (ddprof)
Note: Profiling replicas may still be executing. See the debug dashboard for up to date status.
| Experiment | Variant | Replicates | Failure | Debug Dashboard |
|---|---|---|---|---|
| quality_gate_idle | baseline | 10 | Oom killed | Debug Dashboard |
| quality_gate_idle | comparison | 10 | Oom killed | Debug Dashboard |
| quality_gate_idle_all_features | baseline | 10 | Oom killed | Debug Dashboard |
| quality_gate_idle_all_features | comparison | 10 | Oom killed | Debug Dashboard |
| quality_gate_logs | baseline | 10 | Oom killed | Debug Dashboard |
| quality_gate_logs | comparison | 10 | Oom killed | Debug Dashboard |
| quality_gate_metrics_logs | baseline | 10 | Oom killed | Debug Dashboard |
| quality_gate_metrics_logs | comparison | 10 | Oom killed | Debug Dashboard |
CI Pass/Fail Decision
✅ Passed. All Quality Gates passed.
- quality_gate_logs, bounds check intake_connections: 10/10 replicas passed. Gate passed.
- quality_gate_logs, bounds check missed_bytes: 10/10 replicas passed. Gate passed.
- quality_gate_logs, bounds check total_bytes_received: 10/10 replicas passed. Gate passed.
- quality_gate_logs, bounds check memory_usage: 10/10 replicas passed. Gate passed.
- quality_gate_idle, bounds check memory_usage: 10/10 replicas passed. Gate passed.
- quality_gate_idle, bounds check total_bytes_received: 10/10 replicas passed. Gate passed.
- quality_gate_idle, bounds check intake_connections: 10/10 replicas passed. Gate passed.
- quality_gate_metrics_logs, bounds check memory_usage: 10/10 replicas passed. Gate passed.
- quality_gate_metrics_logs, bounds check missed_bytes: 10/10 replicas passed. Gate passed.
- quality_gate_metrics_logs, bounds check intake_connections: 10/10 replicas passed. Gate passed.
- quality_gate_metrics_logs, bounds check cpu_usage: 10/10 replicas passed. Gate passed.
- quality_gate_metrics_logs, bounds check total_bytes_received: 10/10 replicas passed. Gate passed.
- quality_gate_idle_all_features, bounds check memory_usage: 10/10 replicas passed. Gate passed.
- quality_gate_idle_all_features, bounds check intake_connections: 10/10 replicas passed. Gate passed.
- quality_gate_idle_all_features, bounds check total_bytes_received: 10/10 replicas passed. Gate passed.
1 participant
What does this PR do?
Stamps
ExperimentalAdaptiveSampling.Enabled = falseon every source that enters thelogssource(anomaly detection) parallel pipeline, preventing the global adaptive sampling flag from silently dropping logs before the observer sees them.When
logs_config.experimental_adaptive_sampling.enabledistrueglobally, the decoder'sresolveAdaptiveSamplerEnabledfalls back to that flag for any source withExperimentalAdaptiveSampling == nil. Sources created bylogssourcenever set this field, so they were inheriting anAdaptiveSamplerand dropping noisy log patterns beforeobserver.ObserveLog()— causing the anomaly detection engine to miss anomalies hidden in those suppressed patterns.The per-source
Enabledpointer takes precedence over the global flag in the decoder, so logssource sources always get aNoopSamplerwhile the main agent pipeline continues to sample normally.Two call sites are covered:
source_provider.handleSet— generic workloadmeta container sourcesadSourceManager.AddSource— all AD-scheduled sourcesMotivation
Describe how you validated your changes
Unit tests added in
source_provider_test.goandad_source_manager_test.goasserting thatExperimentalAdaptiveSampling.Enabled == falseon every source entering the pipeline. Full package test suite passes (go test -tags test ./comp/anomalydetection/logssource/impl/).Additional Notes
Made with Cursor