Skip to content

separate step resources - #40

Merged
yashvardhannanavati merged 2 commits into
release-engineering:mainfrom
yashvardhannanavati:codex/separate-step-resources
Nov 2, 2026
Merged

yashvardhannanavati merged 2 commits into
release-engineering:mainfrom
yashvardhannanavati:codex/separate-step-resources

Conversation

@yashvardhannanavati

Copy link
Copy Markdown
Collaborator

No description provided.

Signed-off-by: Yashvardhan Nanavati <yashn@bu.edu>
@qodo-for-releng

Copy link
Copy Markdown

PR Summary by Qodo

Assign separate resource requirements to Tekton task steps

⚙️ Configuration changes 🧪 Tests 🕐 10-20 Minutes


AI Description

• Give artifact retrieval a smaller resource request instead of inheriting build-sized requirements.
• Assign build-specific resources to the multi-architecture step for more appropriate cluster
 scheduling.
• Test that resources are defined per step, not shared through the step template.
Diagram

graph TD
  A["Tekton Task"] --> E["Shared step template"] --> B["Artifact retrieval"] --> C["Multi-arch build"] --> D["Image results"]
  E --> C
Loading
High-Level Assessment

Keep resource requirements on the individual steps. Unlike a shared template allocation, this directly reflects the two steps' different workloads without adding scheduling machinery.

Files changed (2) +28 / -14

Tests (1) +16 / -8
test_iib_image_builder_task.pyVerify per-step resource requirements +16/-8

Verify per-step resource requirements

• Replaces checks for shared template resources with a check that none are defined there. Adds exact resource assertions for both task steps.

tests/tekton/test_iib_image_builder_task.py

Other (1) +12 / -6
iib-image-builder-oci-ta.yamlSet resource requirements separately for artifact retrieval and building +12/-6

Set resource requirements separately for artifact retrieval and building

• Removes shared compute resources from the step template. Artifact retrieval now requests 250m CPU and 512Mi memory with a 2Gi memory limit; the build step requests 1 CPU and 4Gi memory with a 15Gi limit.

task/iib-image-builder-oci-ta/iib-image-builder-oci-ta.yaml

@qodo-for-releng

qodo-for-releng Bot commented Nov 2, 2026 •

Copy link
Copy Markdown

Code Review by Qodo


Resolved findings
1. Cluster operators see stale resource needs ✓ Resolved
Description
Removing stepTemplate.computeResources and setting different per-step requests leaves the Resource
requirements section describing a shared 4 CPU request and 16Gi limit. Operators using that section
to plan cluster quotas or diagnose scheduling now receive values that do not describe either step's
configuration.
Agent prompt
The issue below was found during a code review. Follow the provided context and guidance below and implement a solution

## Issue description
The resource documentation still describes the removed shared step template, rather than the new per-step settings.
## Fix Focus Areas
- task/iib-image-builder-oci-ta/README.md[99-102]
- task/iib-image-builder-oci-ta/iib-image-builder-oci-ta.yaml[98-121]
## Recommended Fix
Replace the shared resource description with the actual CPU and memory requests and memory limits for each step, reflecting the final build-step limit.

ⓘ Copy this prompt and use it to remediate the issue with your preferred AI generation tools


2. Large index builds can run out of memory ✗ Dismissed
Description
build-multi-arch now has a 15Gi memory limit instead of the previous 16Gi limit. A
memory-intensive catalog build that peaks between those limits will be terminated during OPM cache
generation or image building, although it previously fit within the task's allowance.
Agent prompt
The issue below was found during a code review. Follow the provided context and guidance below and implement a solution

## Issue description
The build step's memory ceiling falls from 16Gi to 15Gi, so builds that previously fit can be terminated.
## Fix Focus Areas
- task/iib-image-builder-oci-ta/iib-image-builder-oci-ta.yaml[116-121]
- tests/tekton/test_iib_image_builder_task.py[243-248]
## Recommended Fix
Keep the build step's memory limit at 16Gi while assigning the extraction step its separate resources, and update the corresponding test expectation.

ⓘ Copy this prompt and use it to remediate the issue with your preferred AI generation tools



@fullsend-ai-review

fullsend-ai-review Bot commented Nov 2, 2026 •

Copy link
Copy Markdown

🤖 Finished Review · ✅ Success · Started 3:51 PM UTC · Completed 4:07 PM UTC

Commit: 12c8ad4 · View workflow run →

Runtime: claude · Model: opus → claude-opus-4-6 · Effort: high · Cost: $3.94

Signed-off-by: Yashvardhan Nanavati <yashn@bu.edu>

Assisted-by: Codex
@yashvardhannanavati
yashvardhannanavati force-pushed the codex/separate-step-resources branch from 12c8ad4 to 0a1cdcc Compare November 2, 2026 15:55
@fullsend-ai-review fullsend-ai-review Bot added the risk/low PR risk: low label Nov 2, 2026
@fullsend-ai-review

fullsend-ai-review Bot commented Nov 2, 2026 •

Copy link
Copy Markdown

<!-- fullsend:risk-assessment -->
Risk Assessment: low (1/5)

Details

Small, focused configuration change (3 files, 49 lines) with no protected paths, security-sensitive files, CI workflow, or dependency changes; older files show zero churn and no regression history.

Previous run

<!-- fullsend:risk-assessment -->
Risk Assessment: low (1/5)

Details

Small, focused PR (2 files, 49 net lines) touching only a Tekton YAML and its corresponding test file, with no security-sensitive paths, no CI or dependency changes, and adequate test coverage; stable files with no recent churn keep the overall risk low despite the 5-month code age signal.

@fullsend-ai-review

fullsend-ai-review Bot commented Nov 2, 2026 •

Copy link
Copy Markdown

Review

Findings

Medium

  • [edge-case] task/iib-image-builder-oci-ta/iib-image-builder-oci-ta.yaml:118 — The build-multi-arch step's memory limit was reduced from 16Gi to 15Gi, and the CPU request from 4 cores to 1 core. Workloads that previously consumed between 15Gi and 16Gi of memory will now be OOM-killed. The CPU reduction may affect scheduling on busy clusters, though no CPU limit is set so bursting is possible.
    Remediation: Confirm production workloads do not use between 15Gi and 16Gi of memory. If validated, no change is needed; otherwise keep the memory limit at 16Gi.

  • [stale-doc] README.md:162 — The root README describes TestStepTemplate as verifying "Memory limit, CPU request" but this PR replaced those tests with one asserting computeResources is absent from stepTemplate. The description now contradicts actual test coverage.
    Remediation: Update the TestStepTemplate row to reflect that the class now verifies computeResources is absent from stepTemplate.

Low

  • [missing-authorization] task/iib-image-builder-oci-ta/iib-image-builder-oci-ta.yaml — This PR makes a structural change to compute resource allocation with no linked issue and an empty PR body. The motivation for the new values (CPU reduced from 4 to 1, memory limit from 16Gi to 15Gi) is not documented.
    Remediation: Add a PR body explaining the motivation for the new resource values.

  • [pattern-inconsistency] task/iib-image-builder-oci-ta/iib-image-builder-oci-ta.yaml:104 — CPU value quoting differs between steps: cpu: 250m (unquoted) vs cpu: "1" (quoted). This has a functional reason — YAML parses 250m as a string but bare 1 as an integer — so the inconsistency is aesthetic.
    Remediation: Optionally quote 250m as "250m" for visual consistency.

  • [incomplete-doc] README.md:163 — The root README describes TestTaskSteps coverage but omits the two new per-step compute-resource assertions added by this PR.
    Remediation: Update the TestTaskSteps description to include per-step compute resources.


Next steps:

  • /fs-fix — agent addresses review findings automatically
  • /fs-fix <your instruction> — agent fixes with your specific guidance
  • Push commits directly — review re-runs automatically on push
  • /fs-fix-stop — disable automatic fix runs for this PR
Previous run

Review

Reason: stale-head

The review agent reviewed commit 12c8ad4dda6f4bfe63e85d34d516454a905f47ec but the PR HEAD is now 0a1cdcc6604c955a36dca23bac176e45c80ea919. This review was discarded to avoid approving unreviewed code.

@fullsend-ai-review

Copy link
Copy Markdown

/fs-review

@fullsend-ai-review

fullsend-ai-review Bot commented Nov 2, 2026 •

Copy link
Copy Markdown

🤖 Finished Review · ✅ Success · Started 4:09 PM UTC · Completed 4:28 PM UTC

Commit: 0a1cdcc · View workflow run →

Runtime: claude · Model: opus → claude-opus-4-6 · Effort: high · Cost: $4.34

@fullsend-ai-review fullsend-ai-review Bot left a comment

Copy link
Copy Markdown

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

See the review comment for full details.

@yashvardhannanavati
yashvardhannanavati merged commit 300e74d into release-engineering:main Nov 2, 2026
33 of 35 checks passed
@fullsend-ai-retro

fullsend-ai-retro Bot commented Nov 2, 2026 •

Copy link
Copy Markdown

🤖 Finished Retro · ✅ Success · Started 4:51 PM UTC · Completed 5:02 PM UTC

Commit: 0a1cdcc · View workflow run →

Runtime: claude · Model: opus → claude-opus-4-6 · Effort: high · Cost: $4.01

@fullsend-ai-retro

Copy link
Copy Markdown

Retro: PR #40 — Separate Tekton step resource requirements

Timeline: Human-authored PR (branch codex/separate-step-resources, assisted by Codex) opened at 15:49 UTC on 2026-11-02. Changed 3 files (+34/−15) to move computeResources from the shared stepTemplate to per-step definitions with differentiated profiles.

  • 15:49 — PR #40 opened by yashvardhannanavati. Empty PR body, no linked issue.
  • 15:50 — Qodo bot posted 2 findings: memory limit reduction (16Gi→15Gi) and stale docs. Author replied "It is intentional" and Qodo dismissed.
  • ~15:51 — First fullsend review run (41475382305) dispatched on commit 12c8ad4. Cost: $3.94.
  • 16:07 — First review completed but discarded as stale-head (PR HEAD had moved to 0a1cdcc). Automatic re-dispatch triggered.
  • ~16:09 — Second fullsend review run (41476451364) dispatched on correct HEAD 0a1cdcc. Cost: $4.34.
  • 16:27 — Human reviewer lipoja approved.
  • 16:28 — Second review completed, posted CHANGES_REQUESTED with 5 findings: 2 medium (edge-case memory reduction, stale root README docs), 3 low (missing PR body, CPU quoting inconsistency, incomplete README docs).
  • 16:45 — Human reviewer asergienk approved.
  • 16:49 — PR merged without addressing the root README documentation staleness.

Total review agent cost: $8.28 ($3.94 wasted on stale-head + $4.34 effective).

Key observations

  1. Review agent caught a valid documentation gap that was merged unaddressed. The root README.md test-coverage table (line 162) still says TestStepTemplate covers "Memory limit, CPU request" but the tests now verify computeResources is absent. Filed as the proposal below.

  2. Review agent repeated a finding already dismissed in a prior review tool thread. Qodo flagged the 16Gi→15Gi memory reduction at 15:50, the author dismissed it as intentional at 15:50, then fullsend re-raised the same concern at 16:28 without awareness of the dismissal. New evidence for existing open issues: fullsend-ai/fullsend#4682 and fullsend-ai/fullsend#1672 (suppress already-dismissed findings), fullsend-ai/fullsend#664 (incorporate existing human and third-party bot reviews into assessment), and fullsend-ai/fullsend#4107 (overlapping fullsend/qodo findings).

  3. Stale-head discard wasted $3.94. The first review ran to completion before detecting the HEAD mismatch. New evidence for fullsend-ai/fullsend#1331 (reduce wasted runs from stale-head races) and fullsend-ai/fullsend#2399 (remove stale-head re-dispatch).

  4. PR merged despite CHANGES_REQUESTED verdict. The second review posted at 16:28, one minute after the first human approval; the PR was merged 21 minutes later without addressing findings. New evidence for fullsend-ai/fullsend#7616 (merge can complete before review findings are addressed).

  5. Workflow quality was otherwise good. Risk assessment was accurate (low, 1/5), sub-agent dispatch was well-calibrated for the PR's scope, and the docs-currency finding about the root README was genuinely valuable and correctly categorized.

Proposals filed

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

risk/low PR risk: low

Projects

None yet

Development

Successfully merging this pull request may close these issues.

3 participants