Using Elastic Recheck

[ English | Indonesia | 한국어 (대한민국) | Deutsch | English (United Kingdom) ]

Using Elastic Recheck


This section assumes you have completed Checking Status in Zuul

Allows you to:

  • Enhance the automatic testing the community does on every patch submitted to gerrit
  • Report recurring bugs so that you don’t need to manually ‘recheck’

What to do if a test job fails

When you submit a patch to gerrit and zuul returns the test results for the jobs it ran, sometimes one of those tests fails. Most of the time this indicates there is an issue with your proposed change and the tests are catching it. Sometimes the test run might have tripped over an underlying pre-existing bug in OpenStack. Additionally, some times the infrastructure for running the tests might have had a failure. To figure this out you’ll always need to look at the logs from the failed job to understand what’s happening.

What is elastic-recheck

elastic-recheck is a tool used to track failures in test jobs. It keeps a repository of fingerprints for known bugs that are affecting jobs in the gate. It is then used to both track the rates those bugs are being found and also to leave comments in gerrit and in IRC when it has found a known bug fingerprint in a failure.

elastic-recheck is built on top of an ELK (Elastic Search, Logstash, Kibana) stack where we use Logstash to store all logs from CI jobs in an Elastic Search cluster. We also host a Kibana dashboard which can be used to run queries on the cluster and interacts with the data. elastic-recheck queries the elastic-search cluster for the fingerprints.

You can see the current status of the bugs being tracked by elastic recheck at: Elastic Recheck


Each graph shows how many matches were found for that fingerprint over the past 10 days. It also provides a link to both the launchpad page for the bug, and the kibana dashboard for the underlying elastic-search query used for the fingerprint.

elastic-recheck also has a page to show how many failures were encountered that do not have a matching fingerprint. Typically the more failures that go uncategorized the more unstable the gate is (and OpenStack as a whole).

You can find these pages at:

Unclassified failed integrated gate jobs and Unclassified failed jobs

depending on which jobs you’re interested in.


If you’re interested in more of the theory and history behind the project, this talk from the Juno OpenStack Summit provides a good overview: Elastic Recheck - Tools for Finding Race Conditions in OpenStack

Tracking a new bug in elastic-recheck

When you encounter a failure that’s not being tracked by elastic-recheck and you’ve looked through the logs to determine that it’s not being caused by the proposed change and is affecting other changes you can propose a new elastic-recheck fingerprint.

This guide won’t go into the details of tracing through the logs of a run and finding a good fingerprint, since that’s quite involved, dependent on the job you’re looking at, and already documented in a few places including:

Once you’ve identified a message in the logs that can be used for fingerprinting you need to turn that into an elastic-search query. You can use any of the existing fingerprints as an example: openstack-infra/elastic-recheck

You should also check any elastic search queries using kibana at: Logstash Search

Once you’ve constructed a query and checked in on elastic-search you should create a yaml file in the queries directory of the elastic-recheck git repo. The file name is the launchpad bug number for the bug and the contents are the elastic-search query.

Creative Commons Attribution 3.0 License

Except where otherwise noted, this document is licensed under Creative Commons Attribution 3.0 License. See all OpenStack Legal Documents.