Analyzing Jenkins Jobs

Queue a Jenkins build when you already know the job name and build number and want a report you can review instead of digging through Jenkins by hand. The main choices are whether to wait for the build to finish, how much artifact context to include, and whether to enrich product-bug findings with Jira matches.

Prerequisites

• A running JJI server, and jji pointed at it with --server, JJI_SERVER, or ~/.config/jji/config.toml
• A Jenkins job name and build number
• Jenkins access configured on the server or supplied on the command line
• An AI provider/model available on the server, or chosen per run with --provider and --model
• Optional Jira URL, credentials, and project key if you want Jira enrichment
• Need the initial setup first? See Running Your First Analysis

Quick Example

jji analyze --job-name my-job --build-number 42
jji results show <job-id>

Job queued: <job-id>
Status: queued
Poll: /results/<job-id>

The first command queues the analysis and prints the poll URL. The second command shows the current summary while the run is queued, waiting on Jenkins, running, or complete.

Step-by-Step

1. Queue the Jenkins build.

jji analyze --job-name my-job --build-number 42

If your job lives inside folders, keep the Jenkins path style, such as folder/job-name. If you do not keep a default server in config, prefix the command with jji --server http://localhost:8000.

1. Check what state the queued run is in.

jji results show <job-id>
jji results dashboard

1. Choose whether JJI should wait for Jenkins before it analyzes the build.

jji analyze --job-name my-job --build-number 42 --wait
jji analyze --job-name my-job --build-number 42 --no-wait
jji analyze --job-name my-job --build-number 42 --wait --poll-interval 5 --max-wait 30

Note: The built-in wait defaults are --wait, a 2-minute poll interval, and --max-wait 0, which means "wait as long as needed."

If Jenkins is not configured for the run, JJI skips the waiting step even when wait is enabled.

Warning: --no-wait is best when the Jenkins build is already finished, or when you intentionally want a snapshot of whatever Jenkins exposes at that moment.

1. Override Jenkins or AI settings for just this run.

jji analyze \
  --job-name my-job \
  --build-number 42 \
  --provider gemini \
  --model gemini-2.5-pro \
  --jenkins-url https://jenkins.local \
  --jenkins-user admin \
  --jenkins-password <token> \
  --no-jenkins-ssl-verify

Use this when the run needs different Jenkins access, a different model, or different SSL behavior than your usual defaults. Command-line flags override values from your environment or ~/.config/jji/config.toml for that one run only.

1. Decide how much build-artifact context to include.

jji analyze --job-name my-job --build-number 42 --no-get-job-artifacts
jji analyze --job-name my-job --build-number 42 --get-job-artifacts --jenkins-artifacts-max-size-mb 50

Artifact downloads are enabled by default. When available, JJI downloads build artifacts, unpacks tar and zip archives when possible, and uses that material as extra context for the analysis.

Note: The default artifact size cap is 500 MB. Oversize or failed downloads are skipped instead of failing the whole analysis.

1. Turn Jira enrichment on or off for the run.

jji analyze \
  --job-name my-job \
  --build-number 42 \
  --jira \
  --jira-url https://jira.example.com \
  --jira-project-key PROJ \
  --jira-email user@example.com \
  --jira-api-token <api-token> \
  --jira-max-results 10

Use --no-jira when you want a clean run with no Jira lookups, even if Jira is normally enabled for your server profile. Jira enrichment happens after analysis, and matches are attached only to failures classified as product bugs.

Tip: If you omit both --jira and --no-jira, JJI follows its configured default behavior. In practice, Jira matches only appear when the Jira URL, credentials, and project key are all available.

Advanced Usage

jji results show --full <job-id>
jji analyze --job-name my-job --build-number 42 --tests-repo-url https://github.com/org/tests:feature/bar

The full results view is useful when you want to confirm the exact stored request parameters for a run, including wait settings, provider/model, artifact settings, and Jira toggles. Appending :branch-or-tag to --tests-repo-url lets you pin analysis context to a specific ref for a single run.

For self-signed endpoints, use --no-jenkins-ssl-verify or --no-jira-ssl-verify only on the runs that need them. That keeps your normal defaults strict while still letting one-off runs succeed.

Pipeline builds are supported too. If the parent job fails because child jobs failed, the finished report includes child-job analyses even when the parent has no direct failed tests of its own.

If Jenkins has no structured test report, JJI falls back to console-based analysis. Peer review flags do not apply in that console-only path; for that workflow, see Adding Peer Review with Multiple AI Models.

If you want to analyze failures without a live Jenkins build, see Analyzing JUnit XML and Raw Failures.

Troubleshooting

• Error: No server specified. Point jji at a server with --server, set JJI_SERVER, or add a default server to ~/.config/jji/config.toml.
• Job not found or an immediate Jenkins 404 while waiting. Double-check --job-name and --build-number; jobs inside folders must use folder/job-name.
• Timed out waiting for Jenkins job .... Increase --max-wait, use --max-wait 0 for no time limit, or switch to --no-wait.
• The run completed with no failures. The Jenkins build finished successfully, so JJI returned a completed result with nothing to analyze.
• Artifact evidence is missing. Make sure artifact downloads are enabled and consider raising --jenkins-artifacts-max-size-mb.
• No Jira matches appeared. Confirm Jira was enabled for the run, the Jira URL and project key are set, and valid Cloud or Server/DC credentials were supplied. Remember that Jira matches are only attached to product bug results.

Related Pages

• Running Your First Analysis
• Monitoring and Re-Running Analyses
• Improving Analysis with Repository Context
• Adding Peer Review with Multiple AI Models
• Analyzing JUnit XML and Raw Failures