如何从PostgreSQL获取Ansible Tower/AWX特定任务的STDOUT与STDERR

Absolutely, this is totally feasible—and I’ve dug into this exact scenario before when I needed to pull task output for debugging and audit comments. Your approach is solid; Tower/AWX stores every bit of job and task execution data in its PostgreSQL backend, you just need to map the right tables and relationships.

核心数据表关联逻辑 #

Here’s the breakdown of the key tables you’ll need to query:

• job_events: This is the goldmine for task-level output. Every individual task run (success, failure, skipped, etc.) gets an entry here, with stdout and stderr fields holding the raw output. Depending on your AWX/Tower version, these might be stored as plain text or JSONB (for structured output like stdout_lines).
• jobs: Stores metadata about the entire playbook run (job) — things like job name, status, start/end time. Link this to job_events via the job_id foreign key.
• tasks: Holds the static definition of tasks from your playbooks (name, UUID, etc.). You can link this to job_events using task_id if you need to target tasks by their unique ID instead of name.

示例SQL查询 #

Here’s a practical query to pull the stdout/stderr for a specific task in a specific job:

SELECT
    je.id AS event_id,
    j.name AS job_name,
    je.task_name,
    -- If your stdout is stored as JSON, use this to extract raw text:
    -- je.stdout::json->>'stdout' AS stdout_text,
    je.stdout,
    je.stderr,
    je.started,
    je.finished
FROM
    job_events je
INNER JOIN
    jobs j ON je.job_id = j.id
WHERE
    -- Filter by your target job name
    j.name = 'Deploy Web App'
    -- Filter by your specific task name
    AND je.task_name = 'Install Nginx'
    -- Optional: Filter by event type (success, failure, etc.)
    AND je.event IN ('runner_on_ok', 'runner_on_failure')
ORDER BY
    je.started DESC;

Note: If you have duplicate task names across playbooks, use je.task_id instead of task_name for precise targeting — you can get the task ID from the tasks table or from the AWX UI’s task details page.

替代方案：AWX/Tower API #

While direct DB queries work, using the official API might be safer (especially in production) and avoids dealing with schema changes between versions. For example:

# Replace with your AWX URL, credentials, job ID, and task name
curl -u "admin:your-password" "https://awx.example.com/api/v2/jobs/456/job_events/?task_name=Install Nginx" | jq '.results[] | {task_name, stdout, stderr, started}'

This returns structured JSON with exactly the data you need, no SQL required.

Final Verdict #

Your original idea is absolutely valid. Direct PostgreSQL access is a reliable way to extract task-level stdout/stderr, as long as you map the job_events table to the parent jobs (and optionally tasks) table. Just double-check your AWX/Tower version’s schema documentation if you run into minor discrepancies — older versions might have slightly different table names (like task_results instead of job_events in very old Tower releases).

内容的提问来源于stack exchange，提问作者MMT