throwaway: falsify collapse-settings self-stall CI impact — do not merge

[matthyx]

Manual validation PR — NOT for merge, will be closed after data collection.

kubescape/storage has a confirmed root cause (see storage repo's .omc/plans/collapse-settings-self-stall.md): whenever no CollapseConfiguration/default CR is present, a consolidation-path settings-cache refresh (NewCRDCollapseSettingsProvider / CollapseSettings) can self-stall a goroutine on its own held SQLite write lock for up to the full busy timeout (60s in production). During that window every other writer (every node-agent TS Create, every other consolidation worker) blocks behind it. This fires on essentially every consolidation tick that has data, whenever no CR exists — deterministic, not a race. Test_34_NetworkNeighborsCIDRCollapse is the only existing test that applies this CR today, and only for its own duration.

This PR is the "cheapest falsification experiment" the root-cause doc calls for: adds tests/chart/artifacts/collapseconfiguration-default.yaml (the compiled-in defaults, copied verbatim from storage's artifacts/collapseconfiguration-default-sample.yaml) and a workflow step that applies it right after the chart install, for every test in the matrix — not just Test_34. No storage-image override; uses whatever image storage-tag.sh resolves by default, so this isolates the effect of the CR's presence alone.

Comparison point: recent same-day baselines (no CR, storage main pre-Lane-0) measured 15/16/19 failed out of 31 jobs across 3 runs (mean 16.7) — see #957/#960/#961 and the summary comment on kubescape/storage#399.

If this run (or a couple of reruns) shows a sharp drop in failure count, that's strong evidence this self-stall explains a meaningful chunk of this suite's historical flakiness — independent of, and orthogonal to, the Lane 0 write-path work. If not, its CI-level contribution is small.

Session: https://claude.ai/code/session_01LCMGT6Po2tSr1VEDVrbbYd

🤖 Generated with Claude Code

AI-skills: ralplan,plan | cmds: /compact,/usage-credits

Summary by CodeRabbit

• Tests
  • Added automated test setup that applies a default collapse configuration during component tests.
  • Added configuration coverage for collapse thresholds, path prefixes, network IP groups, and CIDR settings.
  • The workflow retries configuration application and verifies the applied resource.

[coderabbitai]

ℹ️ Recent review info

⚙️ Run configuration

Configuration used: defaults

Review profile: CHILL

Plan: Advanced

Run ID: 793fca25-01bf-444f-bc74-476ac28804c0

📥 Commits

Reviewing files that changed from the base of the PR and between 349c642 and 297a7f8.

📒 Files selected for processing (2)

• .github/workflows/component-tests.yaml
• tests/chart/artifacts/collapseconfiguration-default.yaml

Included review availability: Your plan provides up to 1 included review per hour; 0 remain after this review.

---

📝 Walkthrough

Walkthrough

The component-tests workflow now applies a default CollapseConfiguration custom resource after the node-agent pod is ready. The new artifact defines collapse and network thresholds for the test environment.

Changes

CollapseConfiguration CI setup

Layer / File(s)	Summary
Configure and apply CollapseConfiguration tests/chart/artifacts/collapseconfiguration-default.yaml, .github/workflows/component-tests.yaml	Defines the default CollapseConfiguration resource with collapse and network thresholds. The workflow retries kubectl apply and then prints the applied resource.

Priority: ⬇️ Low

Estimated code review effort: 1 (Trivial) | ~5 minutes

Merge Risk: ⚪ Minimal · up to 297a7

Component-test runs now apply a default CollapseConfiguration after the node-agent is ready, using a resource contract already exercised by component tests. No actionable merge-blocking risk is established by the supplied change context.

Suggested reviewers: entlein

🚥 Pre-merge checks | ✅ 5

✅ Passed checks (5 passed)

Check name	Status	Explanation
Description Check	✅ Passed	Check skipped - CodeRabbit’s high-level summary is enabled.
Title check	✅ Passed	The title clearly identifies the throwaway experiment and its purpose: testing the CI impact of a collapse-settings self-stall. It also correctly states that the change must not be merged.
Docstring Coverage	✅ Passed	No functions found in the changed files to evaluate docstring coverage. Skipping docstring coverage check. Docstring coverage is scoped to functions touched by this diff. Analyzed 0 functions across 0…
Linked Issues check	✅ Passed	Check skipped because no linked issues were found for this pull request.
Out of Scope Changes check	✅ Passed	Check skipped because no linked issues were found for this pull request.

✨ Finishing Touches

🧪 Generate unit tests (beta)

• Create PR with unit tests
• Commit unit tests in branch throwaway/falsify-collapse-stall

---

Thanks for using CodeRabbit! It's free for OSS, and your support helps us grow. If you like it, consider giving us a shout-out.

❤️ Share

• X
• Mastodon
• Reddit
• LinkedIn

_{Comment @coderabbitai help to get the list of available commands.}

[github-actions]

Performance Benchmark Results

Node-Agent Resource Usage

Metric	BEFORE	AFTER	Delta
Avg CPU (cores)	0.216	0.204	-5.4%
Peak CPU (cores)	0.229	0.216	-5.6%
Peak CPU p95 (cores)	0.228	0.214	-6.0%
Avg Memory (MiB)	371.967	320.629	-13.8%
Peak Memory (MiB)	374.949	331.758	-11.5%

Dedup Effectiveness

No data available.