cassandra-commits mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From "Philip Thompson (JIRA)" <j...@apache.org>
Subject [jira] [Resolved] (CASSANDRA-9007) Run stress nightly against trunk in a way that validates
Date Wed, 27 Apr 2016 18:43:13 GMT

     [ https://issues.apache.org/jira/browse/CASSANDRA-9007?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]

Philip Thompson resolved CASSANDRA-9007.
----------------------------------------
    Resolution: Fixed

A series of open source Jepsen tests written by Joel Knighton fulfill what we wanted here.
https://github.com/riptano/jepsen

> Run stress nightly against trunk in a way that validates
> --------------------------------------------------------
>
>                 Key: CASSANDRA-9007
>                 URL: https://issues.apache.org/jira/browse/CASSANDRA-9007
>             Project: Cassandra
>          Issue Type: Task
>            Reporter: Ariel Weisberg
>            Assignee: Philip Thompson
>              Labels: monthly-release
>
> Stress has some very basic validation functionality when used without workload profiles.
It found a bug on trunk when I first ran it so it has value even though the validation is
basic.
> As a beachhead for the kind of blackbox validation that we are missing we can start by
running stress nightly or 24/7 in some rotation.
> There should be two jobs. One job has inverted success criteria (C* should lose some
data) and the job should only "pass" if the failure is detected. This is just to prove that
the harness reports failure if failure occurs.
> Another would be the real job that runs stress, parses and parses the output for reports
of missing data.
> This job is the first pass and basis of what we can point to when a developer makes a
change, implements a feature, or fixes a bug, and say "go add validation to this job."
> Follow on tickets to link to this
> * Test multiple configurations
> * Get stress to validate more query functionality and APIs (counters, LWT, batches)
> * Parse logs and fail tests on error level logs (great way to improve log messages over
time)
> * ?
> I am going to hold off on creating a ton of issues until we have a basic version of the
job running.



--
This message was sent by Atlassian JIRA
(v6.3.4#6332)

Mime
View raw message