[
https://issues.apache.org/jira/browse/SAMZA-850?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=15082169#comment-15082169
]
Xinyu Liu commented on SAMZA-850:
---------------------------------
RB: https://reviews.apache.org/r/41912/
In summary, the implementation adds a new validate-yarn-job.sh script to
validate samza yarn job programmatically. It uses YarnClient to retrieve yarn
job status info, and use JMX to read job metrics. The script will exit 0 if
validation succeeds, and exit 1 if validation fails.
> Yarn Job Validation Tool
> ------------------------
>
> Key: SAMZA-850
> URL: https://issues.apache.org/jira/browse/SAMZA-850
> Project: Samza
> Issue Type: Improvement
> Reporter: Xinyu Liu
> Assignee: Xinyu Liu
>
> We've seen a strong need for validating running samza jobs programmatically.
> Consider the following scenarios:
> 1) running samza as a service (as in SAMZA-849), which requires automated job
> validation when upgrading the framework for large number of jobs;
> 2) production monitoring, automated validation can run periodically and
> trigger alerts;
> 3) integration testing, which can use validation tool to check the job status.
> So a tool is needed to provide the checkout of job health. Minimally it will
> check:
> - job submission
> - app running
> - container count
> It should also be able to extend and validate the job metrics through metrics
> system, such as JMX. The user will be able to validate against important
> metrics such as process_count or message_behind_high_watermark.
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
