Composer CI Error Database

Tracks recurring errors from the osbuild-composer nightly CI pipelines. Updated automatically by the status-dashboard CI analysis pipeline via merge requests.

Database file

ci-error-db.json — single JSON file with all known errors.

Top-level structure

{
  "errors": [ ... ],
  "next_id": 5
}

next_id is the next sequential number for COMPOSER-NNNN ID assignment. Managed by the update script, do not edit manually.

Error entry format

{
  "id": "COMPOSER-0001",
  "signal": "Cannot download repomd.xml: Connection timed out after 30001 milliseconds",
  "description": "RPM repo metadata download times out during osbuild stage",
  "category": "EXTERNAL",
  "component": "org.osbuild.rpm",
  "first_seen": "2026-06-10",
  "last_seen": "2026-06-17",
  "occurrences": [
    {
      "date": "2026-06-10",
      "pipeline_id": 2589716081,
      "job_ids": [14891232513, 14891232309],
      "group": "RHEL-9 Nightly"
    }
  ],
  "resolution": null
}

Fields

Field	Description
id	Unique identifier, COMPOSER-NNNN format. Auto-assigned, never reused.
signal	Verbatim error text from the CI job log. Used for matching against future failures.
description	Short human-readable summary (one sentence).
category	One of: FLAKE, EXTERNAL, BUG, INFRASTRUCTURE.
component	osbuild stage name (e.g. org.osbuild.rpm) or test script name, or null.
first_seen	Date (YYYY-MM-DD) when this error was first observed.
last_seen	Date of the most recent occurrence.
occurrences	Append-only list of each time this error was seen.
resolution	null until resolved. Fill in manually with a short note (e.g. "Fixed in osbuild-composer#5200").

Categories

• FLAKE -- non-deterministic failure (timing, network, resource contention)
• EXTERNAL -- external dependency or service changed behavior or is down
• BUG -- genuine issue in osbuild-composer code or test infrastructure
• INFRASTRUCTURE -- CI runner or environment failure (Docker, OOM, network)

How it works

1. The status-dashboard daily CI analysis fetches failed job logs and the current ci-error-db.json.
2. Claude analyzes the logs and matches failures against known entries or flags new ones.
3. A script applies the updates, assigns IDs for new entries, and creates an MR in this repo.
4. A human reviews and merges the MR.