[autoscaling/cluster] Use target NodePool as base for replica - #50458
[autoscaling/cluster] Use target NodePool as base for replica#50458jennchenn wants to merge 6 commits into
Conversation
|
🎯 Code Coverage (details) 🔗 Commit SHA: 219c4de | Docs | Datadog PR Page | Give us feedback! |
Files inventory check summaryFile checks results against ancestor 95c0fc54: Results for datadog-agent_7.80.0~devel.git.798.219c4de.pipeline.113222045-1_amd64.deb:No change detected |
Static quality checks✅ Please find below the results from static quality gates Successful checksInfo
31 successful checks with minimal change (< 2 KiB)
|
Regression DetectorRegression Detector ResultsMetrics dashboard Baseline: 95c0fc5 Optimization Goals: ✅ No significant changes detected
|
| perf | experiment | goal | Δ mean % | Δ mean % CI | trials | links |
|---|---|---|---|---|---|---|
| ➖ | docker_containers_cpu | % cpu utilization | -0.23 | [-3.10, +2.65] | 1 | Logs |
Fine details of change detection per experiment
| perf | experiment | goal | Δ mean % | Δ mean % CI | trials | links |
|---|---|---|---|---|---|---|
| ➖ | tcp_syslog_to_blackhole | ingress throughput | +2.06 | [+1.88, +2.25] | 1 | Logs |
| ➖ | quality_gate_logs | % cpu utilization | +1.17 | [+0.18, +2.16] | 1 | Logs bounds checks dashboard |
| ➖ | otlp_ingest_metrics | memory utilization | +0.46 | [+0.31, +0.62] | 1 | Logs |
| ➖ | uds_dogstatsd_20mb_12k_contexts_20_senders | memory utilization | +0.40 | [+0.36, +0.45] | 1 | Logs |
| ➖ | ddot_logs | memory utilization | +0.16 | [+0.10, +0.22] | 1 | Logs |
| ➖ | ddot_metrics_sum_cumulative | memory utilization | +0.13 | [-0.03, +0.29] | 1 | Logs |
| ➖ | docker_containers_memory | memory utilization | +0.06 | [-0.05, +0.17] | 1 | Logs |
| ➖ | file_to_blackhole_500ms_latency | egress throughput | +0.03 | [-0.38, +0.44] | 1 | Logs |
| ➖ | quality_gate_idle | memory utilization | +0.02 | [-0.03, +0.06] | 1 | Logs bounds checks dashboard |
| ➖ | uds_dogstatsd_to_api_v3 | ingress throughput | +0.01 | [-0.20, +0.22] | 1 | Logs |
| ➖ | uds_dogstatsd_to_api | ingress throughput | +0.00 | [-0.20, +0.21] | 1 | Logs |
| ➖ | tcp_dd_logs_filter_exclude | ingress throughput | -0.00 | [-0.09, +0.09] | 1 | Logs |
| ➖ | file_to_blackhole_0ms_latency | egress throughput | -0.01 | [-0.54, +0.52] | 1 | Logs |
| ➖ | file_tree | memory utilization | -0.01 | [-0.06, +0.04] | 1 | Logs |
| ➖ | ddot_metrics | memory utilization | -0.02 | [-0.21, +0.18] | 1 | Logs |
| ➖ | file_to_blackhole_1000ms_latency | egress throughput | -0.02 | [-0.45, +0.41] | 1 | Logs |
| ➖ | ddot_metrics_sum_delta | memory utilization | -0.04 | [-0.23, +0.16] | 1 | Logs |
| ➖ | file_to_blackhole_100ms_latency | egress throughput | -0.05 | [-0.20, +0.10] | 1 | Logs |
| ➖ | quality_gate_idle_all_features | memory utilization | -0.08 | [-0.11, -0.04] | 1 | Logs bounds checks dashboard |
| ➖ | docker_containers_cpu | % cpu utilization | -0.23 | [-3.10, +2.65] | 1 | Logs |
| ➖ | ddot_metrics_sum_cumulativetodelta_exporter | memory utilization | -0.35 | [-0.59, -0.12] | 1 | Logs |
| ➖ | quality_gate_metrics_logs | memory utilization | -0.67 | [-0.92, -0.43] | 1 | Logs bounds checks dashboard |
| ➖ | otlp_ingest_logs | memory utilization | -0.76 | [-0.86, -0.66] | 1 | Logs |
Bounds Checks: ✅ Passed
| perf | experiment | bounds_check_name | replicates_passed | observed_value | links |
|---|---|---|---|---|---|
| ✅ | docker_containers_cpu | simple_check_run | 10/10 | 712 ≥ 26 | |
| ✅ | docker_containers_memory | memory_usage | 10/10 | 246.75MiB ≤ 370MiB | |
| ✅ | docker_containers_memory | simple_check_run | 10/10 | 692 ≥ 26 | |
| ✅ | file_to_blackhole_0ms_latency | memory_usage | 10/10 | 0.16GiB ≤ 1.20GiB | |
| ✅ | file_to_blackhole_0ms_latency | missed_bytes | 10/10 | 0B = 0B | |
| ✅ | file_to_blackhole_1000ms_latency | memory_usage | 10/10 | 0.20GiB ≤ 1.20GiB | |
| ✅ | file_to_blackhole_1000ms_latency | missed_bytes | 10/10 | 0B = 0B | |
| ✅ | file_to_blackhole_100ms_latency | memory_usage | 10/10 | 0.17GiB ≤ 1.20GiB | |
| ✅ | file_to_blackhole_100ms_latency | missed_bytes | 10/10 | 0B = 0B | |
| ✅ | file_to_blackhole_500ms_latency | memory_usage | 10/10 | 0.18GiB ≤ 1.20GiB | |
| ✅ | file_to_blackhole_500ms_latency | missed_bytes | 10/10 | 0B = 0B | |
| ✅ | quality_gate_idle | intake_connections | 10/10 | 3 ≤ 4 | bounds checks dashboard |
| ✅ | quality_gate_idle | memory_usage | 10/10 | 143.44MiB ≤ 147MiB | 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 | 472.49MiB ≤ 495MiB | bounds checks dashboard |
| ✅ | quality_gate_logs | intake_connections | 10/10 | 3 ≤ 6 | bounds checks dashboard |
| ✅ | quality_gate_logs | memory_usage | 10/10 | 175.04MiB ≤ 195MiB | bounds checks dashboard |
| ✅ | quality_gate_logs | missed_bytes | 10/10 | 0B = 0B | bounds checks dashboard |
| ✅ | quality_gate_metrics_logs | cpu_usage | 10/10 | 368.41 ≤ 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 | 377.43MiB ≤ 430MiB | bounds checks dashboard |
| ✅ | quality_gate_metrics_logs | missed_bytes | 10/10 | 0B = 0B | 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".
CI Pass/Fail Decision
✅ Passed. All Quality Gates passed.
- quality_gate_logs, bounds check memory_usage: 10/10 replicas passed. Gate passed.
- quality_gate_logs, bounds check missed_bytes: 10/10 replicas passed. Gate passed.
- quality_gate_logs, bounds check intake_connections: 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 missed_bytes: 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 memory_usage: 10/10 replicas passed. Gate passed.
- quality_gate_idle, bounds check intake_connections: 10/10 replicas passed. Gate passed.
- quality_gate_idle, 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 memory_usage: 10/10 replicas passed. Gate passed.
jennchenn
added
changelog/no-changelog
|
@codex review |
![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: 48c855d4da
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 mergeSlices[T any](rc, target []T) []T { | ||
| if len(rc) > 0 { | ||
| return rc | ||
| } |
There was a problem hiding this comment.
mergeSlices uses len(rc) > 0 to decide whether recommendation values are set, which conflates an omitted slice with an explicitly provided empty slice. In replica mode, if a recommendation intentionally sets taints: [] or startupTaints: [] to clear inherited taints from the target NodePool, this code preserves the target taints instead of applying the clear, so stale taints remain on the Datadog-managed NodePool.
Useful? React with 👍 / 👎.
| if len(rc.Spec.Limits) > 0 { | ||
| merged.Spec.Limits = rc.Spec.Limits |
There was a problem hiding this comment.
The limits merge only applies RC values when len(rc.Spec.Limits) > 0, so an explicit empty limits object from the recommendation is treated as "unset" and the target limits are retained. That prevents replica recommendations from removing existing spec.limits inherited from the target NodePool, which breaks the documented "RC overrides when set" behavior.
Useful? React with 👍 / 👎.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
The replica merge used the target NodePool fetched from the cluster as its base, carrying ResourceVersion/UID/Generation/CreationTimestamp/ ManagedFields into the merged result. Sending that as a Create made the API server reject it with "resourceVersion should not be set on objects to be created". Replace the top-level ObjectMeta wholesale with a fresh struct so server- set fields are zero by construction, and exercise the path in fixture + assertions. Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Add two small generic helpers that express "RC wins if set, else preserve target" without repeating nil/len guards for every field. Replace reflect.DeepEqual zero-checks with explicit field checks (ConsolidationPolicy != "" for Disruption, Duration/Raw for ExpireAfter), dropping the reflect import.
jennchenn
changed the title
48c855d to
219c4de
Compare
There was a problem hiding this comment.
💡 Codex Review
Here are some automated review suggestions for this pull request.
Reviewed commit: 219c4de7e6
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".
| } | ||
|
|
||
| // NodeClaimTemplateSpec fields: use RC values if set, else fallback to target | ||
| merged.Spec.Template.Spec.Requirements = mergeSlices(rc.Spec.Template.Spec.Requirements, merged.Spec.Template.Spec.Requirements) |
There was a problem hiding this comment.
When building replica NodePools, this assignment uses mergeSlices (which checks len(rc) > 0), so a recommendation that intentionally sets spec.template.spec.requirements: [] is treated the same as “unset” and the target NodePool’s requirements are kept. In replica mode that prevents clearing inherited constraints from the target, so stale requirements continue to restrict scheduling instead of applying the recommendation’s explicit empty override.
Useful? React with 👍 / 👎.
|
|
||
| // NodePoolSpec fields: use RC values if set, else fallback to target | ||
| if rc.Spec.Weight != nil { | ||
| merged.Spec.Weight = rc.Spec.Weight |
There was a problem hiding this comment.
What if this is lower than the targetNp? I think we should always check the targetNp weight.
| Labels: maps.Clone(rc.Labels), | ||
| Annotations: maps.Clone(rc.Annotations), |
There was a problem hiding this comment.
Should we not give preference to the targetNp here?
|
|
||
| func GetNodePoolWeight(replicaNp *karpenterv1.NodePool) *int32 { | ||
| // BuildReplicaNodePool produces a NodePool for a Datadog-managed replica of an existing target NodePool | ||
| // The target is used as the base and RC values are applied on top: |
There was a problem hiding this comment.
Should we do a merge? I would think we would prefer the target's values if there is a conflict.
| } | ||
|
|
||
| // NodeClaimTemplateSpec fields: use RC values if set, else fallback to target | ||
| merged.Spec.Template.Spec.Requirements = mergeSlices(rc.Spec.Template.Spec.Requirements, merged.Spec.Template.Spec.Requirements) |
There was a problem hiding this comment.
For requirements, we should check if there are some conflicts that can't be overcome with a merge, like arch/os for example. If they do, then we can't proceed with the merge.
|
This pull request has been automatically marked as stale because it has not had activity in the past 15 days. It will be closed in 30 days if no further activity occurs. If this pull request is still relevant, adding a comment or pushing new commits will keep it open. Also, you can always reopen the pull request if you missed the window. Thank you for your contributions! |
|
No longer necessary as we are now sending the entire spec from the backend |
2 participants
What does this PR do?
When the cluster-agent autoscaling controller produces a Datadog-managed replica from a user-created target NodePool, use the target as the merge base and overlay RC values selectively, instead of building from RC alone.
Spec.Template.Spec.RequirementsNodeClassRef,Disruption,ExpireAfter,Limits,Taints,StartupTaints,TerminationGracePeriod,Replicastarget.Weight + 1Standalone path (no target) is unchanged.
Motivation
Restores merge semantics from #42171. Without this, operator-edited fields on the target silently disappear from the replica because the RC manifest only carries what the recommendation engine knows about.
Describe how you validated your changes
pkg/clusteragent/autoscaling/cluster/model/cover all merge cases.Spec.Weight = target+1.Additional Notes
🤖 Generated with Claude Code