[cristim]

Closes #30

Change

• buildResult (pkg/ladder/allocate.go) calls the new holdReshapedBufferAllocations right after splitGap. When a layer has a reshape and is buffer-only, its allocation becomes an informational ActionHold (rationale names the pending reshape) and the reshape still appears. Other layers and allocations are unchanged.
• Holds in this package carry free-string rationales and there is no reason enum, so no exported enum value or signature changed.
• A layer that also has the base role (the merged Azure reservation) is not held: its allocation is base+buffer and dropping it would also drop the base purchase. This becomes reachable when feat(ladder): phase 4 - Azure (BillingBenefits savings plans + reservation exchange) #28 (Azure ladder) lands; the follow-up is to zero only the buffer share in assembleAllocations (mergedGap = baseGap + bufferGap).
• Two existing max-actions tests used the conflicting buffer+reshape fixture. Their expected truncation holds now exclude ConvertibleRI, which is held earlier by this change.

Tests

• Buffer layer with a pending reshape: no buffer purchase, a Hold, the reshape, other allocations kept.
• No reshape: buffer still bought.
• Reshape on another layer, and the merged Azure layer: allocation kept (helper-level, since valid topologies allow one buffer layer).
• Mutation probe: removing the conversion fails the new test and both updated max-actions tests; reverted.
• go test -race -count=1 ./... in pkg (workspace and GOWORK=off) and providers/aws, and golangci-lint on pkg: clean.

🤖 Generated with Claude Code

[coderabbitai]

Warning

Review limit reached

• Run on-demand review

_{This review includes 3 billable files and costs up to $0.75.}

• Ask an admin to make reviews automatic

Open in CodeRabbit

Reviews can continue after your included limit without a manual trigger. An admin must approve usage-based billing.

Or wait 1 minute for your next included review.

Check out review usage here.

View limit details

Limit details: You’ve used all 2 included reviews currently available. Your 59 included PR review attempts over the past 7 days set your current allowance at 2 reviews per hour.

Learn how review limits work.

Review configuration:

Run configuration

• Source mode: Auto
• Repository: LeanerCloud/cloud-commitments-go
• Branch: fix/30-hold-buffer-on-reshape
• File: .coderabbit.yaml
• Commit: d8f389f
• Layers: Repository YAML
• Review profile: CHILL
• Plan: Essentials
• Run ID: a8434348-78ba-4d1d-9dd6-9f011c23c004

Commits

Reviewing files that changed from the base of the PR and between bdec74d and d8f389f.

Files selected for processing: 3

• CHANGELOG.md
• pkg/ladder/allocate.go
• pkg/ladder/allocate_test.go

• Autofix · Keep fixing CodeRabbit findings and required CI, and resolving merge conflicts

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

[cristim]

Review verdicts at 4a8474c: two independent cold reviews (non-author), both APPROVE.

• Reshapes are emitted only for buffer layers and keyed by LayerType. Validation allows at most one buffer layer, so a reshape cannot miss its buffer or hit the wrong one. In the AWS topology that layer is ConvertibleRI.
• The hold only removes an allocation, so no layer buys above its own gap and total spend never rises. Under a binding MaxHourlyCommitPerRun the cap scale stays <= 1.
• The changed max-actions expectations follow from the fix: the old ones truncated the very ConvertibleRI buffer purchase that feat(ladder): reconcile simultaneous buffer purchase + reshape recommendations in one run #30 describes.
• Mutation probes: removing the call, holding merged layers, matching a reshape on any layer, and not appending the holds each fail tests. go test -race -count=1 passes in pkg (with and without the workspace) and providers/aws. Lint is clean.
• Known limitation: a merged base+buffer layer is not held. No caller builds one today. This is tracked for the Azure ladder (feat(ladder): phase 4 - Azure (BillingBenefits savings plans + reservation exchange) #28).
• Follow-ups, not blocking: a test that pins the hold running before the cap guardrail, and a unit label in the hold rationale.

CodeRabbit did not review (review limit reached).

[claude-cc-libs]