Mismatch between JSON sidecar formats accepted by MetaCat and those required by declad dropbox ingestion

[DouglasLeeTucker]

Summary

During a large‑scale MetaCat stress test (100k files) performed on dtucker@fifeutilgpvm03.fnal.gov, I generated JSON sidecar metadata files that MetaCat accepted without issue. However, when attempting to ingest the same files via declad’s dropbox mechanism on hypotpro@fermicloud848.fnal.gov, declad rejected them.

To proceed, I had to rewrite all JSON sidecars into a different structure — one that declad accepts but MetaCat does not require when ingesting directly.

This indicates a mismatch between:

• the JSON structure MetaCat accepts directly, and
• the JSON structure declad requires before it will ingest and forward metadata to MetaCat.

This mismatch caused ingestion failures and required a script to rewrite the JSON metadata files.

Environment

MetaCat stress test environment
Host: dtucker@fifeutilgpvm03.fnal.gov
MetaCat version: 4.1.?
JSON sidecars accepted directly by MetaCat

declad ingestion environment

• Host: hypotpro@fermicloud848.fnal.gov
• declad version: 2.3.8
• declad dropbox ingestion using /home/hypotpro/declad_848/declad_config.yaml
• declad rejected the original JSON sidecars

Example of the Mismatch
1. JSON sidecar that MetaCat accepted directly
This file (/home/dtucker/WORK/GitHub/MetacatStressTest/python/synthetic_minimal_n100000/data_ffff9d76-0e95-4062-81a5-edd7d0279791.parquet.json.orig) is representative of the structure used during the MetaCat stress test:

{
  "dh.type": "other",
  "fn.configuration": "c20240226",
  "fn.description": "metacat_stress_test_20260216_5",
  "fn.format": "txt",
  "fn.owner": "dtucker",
  "fn.tier": "etc",
  "rs.runs": [
    1000002
  ]
}

MetaCat accepted this structure without requiring additional top‑level fields.

2. Equivalent JSON sidecar required by declad
To make declad ingest the same file, I had to rewrite the JSON into the following structure (see /home/dtucker/WORK/GitHub/MetacatStressTest/python/synthetic_minimal_n100000/data_ffff9d76-0e95-4062-81a5-edd7d0279791.parquet.json:

{
  "name": "data_ffff9d76-0e95-4062-81a5-edd7d0279791.parquet",
  "namespace": "hypotpro",
  "size": 6967,
  "checksums": {
    "adler32": "b9a2d1fc"
  },
  "metadata": {
    "dh.type": "other",
    "fn.configuration": "c20240226",
    "fn.description": "metacat_stress_test_20260216_5",
    "fn.format": "txt",
    "fn.owner": "hypotraw",
    "fn.tier": "etc",
    "rs.runs": [
      1000002
    ]
  }
}

Key differences required by declad:

• Mandatory top‑level fields:
  -- name
  -- namespace
  -- size
  -- checksums
• All metadata must be nested under "metadata"
• Dot‑notation keys must be preserved inside "metadata"
• rs.runs must be a list, not a scalar
• Missing or differently‑named fields cause declad to reject the file silently or stall

Possible Actions (suggested by Microsoft Copilot)

• Document the required JSON structure for declad dropbox ingestion, including:
  -- required top‑level fields
  -- required nesting under "metadata"
  -- required checksum formats
  -- required run‑number fields
• Clarify whether declad should accept the same JSON structure that MetaCat accepts directly, or whether the two systems are intentionally different.
• Improve error reporting when JSON sidecars are malformed or missing required fields.
• (Optional) Add a validation mode to declad that checks JSON sidecars and reports structural issues before ingestion.

(Gory details can be found in this Microsoft Copilot conversation: link )