Repository navigation
perf(aws): populate ComputeDetails VCPU/MemoryGB from EC2 DescribeInstanceTypes catalogue #99
Description
Summary
PR #97 extended common.ComputeDetails with VCPU + MemoryGB fields. The AWS compute path (providers/aws/recommendations/parser_services.go::ec2Info) doesn't populate them because the upstream Cost Explorer EC2InstanceDetails doesn't carry that data — populating requires a new ec2.DescribeInstanceTypes caller + per-region cache + mocks for tests.
Current behaviour
parser_services.go:76 builds &common.ComputeDetails{} from CE data only. The new VCPU / MemoryGB fields ship as zero (omitted from JSON via omitempty).
Steps to verify the gap
Query a deployed environment with AWS EC2 RI recommendations and observe that responses have details.vcpu / details.memory_gb absent.
Expected behaviour
Populated from ec2.DescribeInstanceTypes(InstanceTypes: [m5.large, ...]) keyed by the converter's seen instance types. Cache the catalogue per-region for the duration of a recommendation collection.
Proposed fix
- Add an EC2 instance-type catalogue caller in
providers/aws/services/ec2/— wrapsDescribeInstanceTypes, caches per-region per-collection-run. - Inject the catalogue into
parser_services.go(or the equivalent post-fetch enrichment point). - Populate
VCPUfromInstanceTypeInfo.VCpuInfo.DefaultVCpus,MemoryGBfromInstanceTypeInfo.MemoryInfo.SizeInMiB / 1024. - New mocks in test files for the new SDK call.
References
- PR perf(common): extend ComputeDetails with VCPU + MemoryGB #97 — schema extension (
ComputeDetails.VCPU+MemoryGB) - Issue perf(common): extend ComputeDetails with vCPU + MemoryGB to unlock Azure compute SKU enrichment #82 — parent
providers/aws/recommendations/parser_services.goproviders/aws/services/ec2/client.gopkg/common/types.go::ComputeDetails
Severity
Low — same as Azure side; UI degrades to "unknown".
Effort
Medium — net-new SDK caller + per-region cache + per-collection lifecycle + mocks. Larger than the Azure follow-up.
Activity
cristim commentedon Apr 28, 2026
P3 — graceful degradation to 'unknown' in the UI, low urgency. EC2 VCPU/MemoryGB population requires a new
DescribeInstanceTypescaller + per-region cache, making this effort/m. Not blocking any shipped feature; defer to after the override cluster (#107, #111) is resolved. (triage agent wave1-B)cristim commentedon May 28, 2026
Verification: cannot-verify-autopilot
PR #229 (merge commit c05c0d1, confirmed in origin/feat/multicloud-web-frontend) closes #148 (Azure compute vCPU/MemoryGB via cached SKU catalogue), not #99. Issue #99 is specifically about the AWS path:
providers/aws/recommendations/parser_services.go::ec2Infoneeds a newec2.DescribeInstanceTypescaller + per-region cache + mocks to populateVCPU/MemoryGB. Verified on feat branch thatproviders/aws/recommendations/has noDescribeInstanceTypescall. Issue remains open.perf(aws/recs): populate ComputeDetails VCPU/MemoryGB via DescribeIns…
perf(aws/recs): populate ComputeDetails VCPU/MemoryGB via DescribeIns…
perf(aws/recs): populate ComputeDetails VCPU/MemoryGB via DescribeIns…
test(aws/recs): cover SizeInMiB-nil + assert no NextPage on canceled …
test(aws/recs): cover SizeInMiB-nil + assert no NextPage on canceled …
test(aws/recs): cover SizeInMiB-nil + assert no NextPage on canceled …
test(aws/recs): cover SizeInMiB-nil + assert no NextPage on canceled …
test(aws/recs): cover SizeInMiB-nil + assert no NextPage on canceled …