[
https://issues.apache.org/jira/browse/HBASE-13187?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=14369705#comment-14369705
]
stack commented on HBASE-13187:
-------------------------------
bq. ....but I suggest to generate tiny and big value based on rowkey(or some
other values in the row) so that we can also verify whether the data of tiny
and big family are also correct?
Would there be any value doing this extra work? The test just ensures
referential integrity between rows no matter what the cluster perturbation. The
patch adds a new 'agitation' making it very likely column families will flush
independently (when flush-columnfamily is enabled). Having the middle 'meta'
column family for checking referenced row is still present is sufficient?
Having 'big' and 'tiny' columns do it independently adds no extra insurance
(that I can see). Maybe we could have 'big' and 'tiny' column families
reference other rows altogether making a sort of mesh of references but if it
is row-based references that we are checking, I just see us doing extra
footwork to no benefit.
A cool 'new' test would do up this mesh-of-references but the references would
be between cells rather than rows.
Your comment does prompt me to add to the verification step a check that the
big and tiny are present on every row.
Thanks [~Apache9]
> Add ITBLL that exercises per CF flush
> -------------------------------------
>
> Key: HBASE-13187
> URL: https://issues.apache.org/jira/browse/HBASE-13187
> Project: HBase
> Issue Type: Task
> Components: integration tests
> Reporter: stack
> Assignee: stack
> Priority: Critical
> Fix For: 2.0.0, 1.1.0
>
> Attachments: 13187.txt
>
>
> Let me work on this. It would be excellent if we could have confidence to
> turn this on earlier rather than later.
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
