baseline-reindex

name	baseline-reindex
description	Re-index the OPCT baseline results stored in S3 and invalidate the CloudFront cache. Use when baseline results are missing, after new results are published, or when the index appears corrupted.
allowed-tools	Bash, Read, WebFetch

Baseline Reindex

Re-index the OPCT baseline results stored in S3 and invalidate the CloudFront cache.

When to use

• After new baseline results have been published to S3
• When the index.json is missing entries or appears corrupted
• After manual S3 object additions/deletions

Prerequisites

• AWS credentials configured (via env vars, default profile, or instance role)
• OPCT_ENABLE_ADM_BASELINE=1 environment variable set

Architecture

The baseline indexer operates on an S3 bucket (opct-archive) under the prefix api/v0/result/summary/.

Object naming convention: {ocpVersion}_{platformType}_{timestamp}.json

• Example: 4.19_None_20260501214947.json

Special objects:

• index.json — master index with all results and latest map
• {version}_{platform}_latest.json — alias copies of the latest result per version/platform

Incremental indexing flow:

1. Lists all S3 objects (paginated via ListObjectsPages)
2. Loads existing index.json from S3
3. Builds a set of already-indexed paths
4. Only fetches and parses metadata for new objects
5. Recalculates latest entries across all results
6. Writes updated index.json and copies latest aliases
7. Invalidates CloudFront cache paths

Falls back to full reindex if no existing index is found.

Key files

• internal/report/baseline/indexer.go — CreateBaselineIndex(), ListObjects(), loadIndexFromS3(), fetchObjectMetadata()
• internal/report/baseline/aws.go — S3 and CloudFront client creation
• internal/report/baseline/baseline.go — BaselineConfig, S3 client setup, API readers
• internal/report/baseline/data.go — BaselineData, metadata extraction from setup.api
• internal/report/baseline/uploader.go — UploadBaseline() for publishing results

Commands

# Build
make build-linux-amd64

# Run the indexer
OPCT_ENABLE_ADM_BASELINE="1" opct adm baseline indexer --log-level=debug

# List baseline results
opct adm baseline list

# Publish a result
OPCT_ENABLE_ADM_BASELINE="1" opct adm baseline publish --log-level=debug <artifact.tar.gz>

S3 bucket structure

opct-archive/
├── uploads/                              # Raw artifact archives
│   └── *.tar.gz
└── api/v0/result/summary/               # Processed summaries
    ├── index.json                        # Master index (~530KB)
    ├── 4.13_None_20240729023635.json     # Individual results (~40KB each)
    ├── 4.13_None_latest.json             # Latest alias (copy of newest)
    └── ...

Configuration

Env var	Default	Description
OPCT_ENABLE_ADM_BASELINE	unset	Must be 1 to enable admin baseline features
OPCT_EXP_BUCKET_NAME	opct-archive	Override S3 bucket name
OPCT_EXP_BUCKET_REGION	us-east-1	Override S3 bucket region
AWS_SHARED_CREDENTIALS_FILE	default	Path to AWS credentials file (used in CI)

CI integration

The indexer runs in OpenShift CI after baseline results are published:

• Step reference: https://steps.ci.openshift.org/reference/opct-conformance-test-results
• Script: opct-conformance-test-results-commands.sh
• Credentials: /var/run/vault/opct/.awscred-vmc-ci-opct-uploader (via AWS_SHARED_CREDENTIALS_FILE)
• CloudFront distribution: E3MJR7MT6EHHJC
• Local clone of CI step registry: check ci-operator/step-registry/opct/ in the openshift/release repo

Known issues and history

• OPCT-399: S3 pagination was capped at 1000 objects, _latest.json files polluted the index, CloudFront credentials required hardcoded profile. Fixed in PR #212.
• Index currently has ~1,156 objects spanning OCP versions 4.13–4.22.