[
https://issues.apache.org/jira/browse/CASSANDRA-8773?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=14378037#comment-14378037
]
Benedict commented on CASSANDRA-8773:
-------------------------------------
Just to elaborate a little on how this works: much like "insert" is a special
kind of operation, "validate" is now another special kind of operation, which
with equal probability runs one of a range of possible queries against the
database for which it can reliably construct the result it should expect. This
set of queries is: a whole partition retrieval, an (inclusive LB, exclusive UB)
query on clustering prefixes, and all four possible inclusive/exclusive LB/UB
on a full clustering column provision. Only actual values present for the
partition are provided for each parameter, and the value is selected at random.
Ultimately we should support arbitrary value provision, including present or
not-present values, and all possible inclusive/exclusive combinations. But this
should suffice to stretch the system more than is currently the case. Its main
problematic constraint is that the dataset be fully populated before starting.
It can support overwrites (so a mixed workload of inserts can be run), but
since the data is only overwriting like data this constraints the usefulness of
the validation. This should all be solved in CASSANDRA-8986.
> cassandra-stress should validate its results in "user" mode
> -----------------------------------------------------------
>
> Key: CASSANDRA-8773
> URL: https://issues.apache.org/jira/browse/CASSANDRA-8773
> Project: Cassandra
> Issue Type: Improvement
> Components: Tools
> Reporter: Benedict
> Assignee: Benedict
> Labels: stress
> Fix For: 2.1.4
>
>
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
