[
https://issues.apache.org/jira/browse/CASSANDRA-9057?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=14514586#comment-14514586
]
Sam Tunnicliffe commented on CASSANDRA-9057:
--------------------------------------------
I don't believe the patch is quite complete as I think the same issue will also
affect indexes on map keys. That is, map entries with an indexed key < 64k but
a non-indexed value > 64k will be incorrectly rejected.
A couple more comments:
* The test you added is a bit hard to parse, at least by me. Using the BatchLog
cf seems weird and re-using the same string ("ABC") for both the partition key
and column name made me have to read it several times before I was sure I
wasn't missing something.
* It would be good to add some tests to IndexedValuesValidationTest, extending
it slightly so we can test scenarios we expect to succeed as well as those we
expect to fail. e.g. an index on a clustering or pk column < 64k with a regular
column > 64k etc
* Related to the above, the patch contains no additional coverage for the index
on partition key component
Nits:
* Don't need the braces in {{CompositesIndexOnClusterKey#validate}}
* It would seem more natural to order the args of {{ThriftValidation#validate}}
like {{metadata, key, scName, column}} for consistency with
{{validateColumnOrSuperColumn}} (sorry for the OCD nit-pickery)
> index validation fails for non-indexed column
> ---------------------------------------------
>
> Key: CASSANDRA-9057
> URL: https://issues.apache.org/jira/browse/CASSANDRA-9057
> Project: Cassandra
> Issue Type: Bug
> Reporter: Eric Evans
> Assignee: Carl Yeksigian
> Fix For: 2.1.5
>
> Attachments: 9057-2.1.txt
>
>
> On 2.1.3, updates are failing with an InvalidRequestException when an
> unindexed column is greater than the maximum allowed for indexed entries.
> {noformat}
> ResponseError: Can't index column value of size 1483409 for index null on
> local_group_default_T_parsoid_html.data
> {noformat}
> In this case, the update _does_ include a 1483409 byte column value, but it
> is for a column that is not indexed, (the single indexed column is < 32
> bytes), presumably this is why
> {{cfm.getColumnDefinition(cell.name()).getIndexName()}} returns {{null}}.
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
