microshift-ci

microshift-ci

Analyze MicroShift CI failures, produce HTML reports, and create JIRA bugs.

Installation

/plugin marketplace add openshift-eng/edge-tooling
/plugin install microshift-ci

CI Doctor Pipeline

The full CI doctor pipeline is driven by run-doctor.py — a deterministic Python script that orchestrates all stages (prepare, graphs, analyze, bugs, finalize). LLM agents are used only for per-job root cause analysis and Jira bug correlation.

python3 plugins/microshift-ci/scripts/run-doctor.py \
    --releases 4.19,4.20,4.21,4.22 --workdir /tmp/workdir

# Include pull request analysis
python3 plugins/microshift-ci/scripts/run-doctor.py \
    --releases 4.19,4.20,4.21,4.22 --workdir /tmp/workdir \
    --pull-requests --repo openshift/microshift

# Run specific stages only
python3 plugins/microshift-ci/scripts/run-doctor.py \
    --releases 4.22 --workdir /tmp/workdir --stages analyze,finalize

Skills

Skill	Description
/microshift-ci:prow-job	Root cause analysis of a single Prow job
/microshift-ci:test-job	Comprehensive job metadata and scenario results
/microshift-ci:test-scenario	Analyze individual test scenario results
/microshift-ci:create-bugs	Search JIRA for duplicates and create bugs (dry-run by default)
/microshift-ci:close-stale-bugs	Close stale, unlinked, unassigned AI-generated bugs (dry-run by default)
/microshift-ci:continue-session	Download CI Doctor artifacts from a completed prow job
/microshift-ci:fix-test-bugs	Attempt to fix CI bugs by opening PRs in openshift/microshift (dry-run by default)

Usage

Full pipeline

python3 plugins/microshift-ci/scripts/run-doctor.py \
    --releases 4.19,4.20,4.21,4.22 --workdir /tmp/workdir

Single job analysis

/microshift-ci:prow-job https://prow.ci.openshift.org/view/gs/test-platform-results/logs/<job-name>/<job-id>

Create bugs from analysis

/microshift-ci:create-bugs 4.22           # dry-run
/microshift-ci:create-bugs 4.22 --create  # create/update bugs

Note: The scheduled CI Doctor job runs create-bugs in dry-run mode only (for bug correlation in the HTML report). Bugs are created manually via the prefilled "Create Bug in JIRA" buttons in the report, or by running create-bugs --create explicitly.

Requirements

• gsutil CLI (uses anonymous access on public GCS buckets)
• gh CLI (authenticated with access to openshift/microshift)
• Jira MCP server configured (for bug correlation)
• Python 3
• pcp-export-pcp2json (for PCP performance graphs)
• matplotlib Python package (for PCP graph plotting)
• Category: ci-cd

Author

ggiguash