baseline-debug

name	baseline-debug
description	Diagnose issues with the OPCT baseline index, S3 storage, or CloudFront distribution. Use when baseline results are missing, stale, or CI publish/indexer steps are failing.
allowed-tools	Bash, Read, WebFetch

Baseline Debug

Diagnose issues with the OPCT baseline index, S3 storage, or CloudFront distribution.

When to use

• Baseline results are missing from opct adm baseline list
• Index appears stale or corrupted
• CloudFront serving outdated content
• CI baseline publish/indexer steps failing

Diagnostic steps

When the user reports baseline issues, follow this sequence:

1. Check S3 bucket contents

# Count total objects
aws s3 ls s3://opct-archive/api/v0/result/summary/ | wc -l

# List objects, save for analysis
aws s3 ls s3://opct-archive/api/v0/result/summary/ > /tmp/bucket-list.txt

# Count by type
grep -c '_latest.json' /tmp/bucket-list.txt   # latest aliases
grep -vc '_latest.json\|index.json' /tmp/bucket-list.txt  # real objects

2. Download and inspect the index

aws s3 cp s3://opct-archive/api/v0/result/summary/index.json /tmp/index.json

# Quick analysis
python3 -c "
import json
with open('/tmp/index.json') as f:
    idx = json.load(f)
print(f'Total results: {len(idx[\"results\"])}')
print(f'Latest entries: {len(idx[\"latest\"])}')
print(f'Last update: {idx[\"date\"]}')
# Check for _latest pollution
polluted = [r for r in idx['results'] if '_latest' in r['name']]
print(f'Polluted _latest entries: {len(polluted)}')
# Check version coverage
from collections import Counter
versions = Counter(r.get('openshift_version','?') for r in idx['results'] if '_latest' not in r['name'])
print('Entries per version:')
for v, c in sorted(versions.items()):
    print(f'  {v}: {c}')
"

3. Compare index vs S3

python3 -c "
import json
# Load index
with open('/tmp/index.json') as f:
    idx = json.load(f)
indexed_names = {r['name'] for r in idx['results'] if '_latest' not in r['name']}

# Load bucket listing
with open('/tmp/bucket-list.txt') as f:
    lines = f.readlines()
s3_names = set()
for l in lines:
    name = l.strip().split()[-1].replace('.json', '')
    if '_latest' not in name and name != 'index':
        s3_names.add(name)

missing = s3_names - indexed_names
extra = indexed_names - s3_names
print(f'In S3 but not in index: {len(missing)}')
for m in sorted(missing)[:20]:
    print(f'  {m}')
print(f'In index but not in S3: {len(extra)}')
"

4. Check CloudFront cache

# Fetch via CloudFront and compare
curl -s https://d23912a6309zf7.cloudfront.net/result/summary/index.json | python3 -c "
import json, sys
idx = json.load(sys.stdin)
print(f'CloudFront index: {len(idx[\"results\"])} results, updated {idx[\"date\"]}')
"

# If stale, invalidate manually
aws cloudfront create-invalidation \
    --distribution-id E3MJR7MT6EHHJC \
    --paths /result/summary/index.json /result/summary/*_latest.json

Common issues

Symptom	Cause	Fix
Index has exactly 1000 results	Pagination bug (pre-OPCT-399)	Update to fixed version, rerun indexer
_latest entries in results array	Filter bug (pre-OPCT-399)	Update to fixed version, rerun indexer (auto-cleans)
CloudFront NoCredentialProviders	Hardcoded AWS profile (pre-OPCT-399)	Update to fixed version
Missing recent versions	Pagination cap or index not rerun	Rerun opct adm baseline indexer
CloudFront shows old data	Cache not invalidated	Manual invalidation (see step 4)

Key constants

• S3 bucket: opct-archive
• S3 region: us-east-1
• CloudFront distribution: E3MJR7MT6EHHJC
• CloudFront URL: https://d23912a6309zf7.cloudfront.net
• Index path: api/v0/result/summary/index.json