[
https://issues.apache.org/jira/browse/CASSANDRA-14913?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16700469#comment-16700469
]
Aleksey Yeschenko commented on CASSANDRA-14913:
-----------------------------------------------
[~ifesdjeen] I wouldn't mind it, personally. The change would be more involved:
would have to add an extra column ({{kind}}) to the schema table in 3.0 (that
already exists in 4.0, making this change super trivial). Also extend
{{DroppedColumn}} a little. But that doesn't sound prohibitively involved to me.
If you are interested in doing the backport, please do. Otherwise I'll do it
eventually.
While at it, might also backport 4.0 prohibition on re-adding columns with a
different type from the dropped one if nobody objects.
> Forbid re-adding static columns as regular and vice versa
> ---------------------------------------------------------
>
> Key: CASSANDRA-14913
> URL: https://issues.apache.org/jira/browse/CASSANDRA-14913
> Project: Cassandra
> Issue Type: Bug
> Components: Distributed Metadata
> Reporter: Aleksey Yeschenko
> Assignee: Aleksey Yeschenko
> Priority: Major
> Fix For: 4.0
>
>
> Re-adding a dropped column with an incompatible kind (dropped regular
> re-added as static, or dropped static re-added as regular) can ultimately
> result in corruption (see CASSANDRA-14843 for more context). In 3.x,
> unfortunately, we don’t persist enough context when dropping a column. In
> trunk, however, we do, and it’s trivial to forbid this operation, as we
> should.
--
This message was sent by Atlassian JIRA
(v7.6.3#76005)
---------------------------------------------------------------------
To unsubscribe, e-mail: commits-unsubscr...@cassandra.apache.org
For additional commands, e-mail: commits-h...@cassandra.apache.org
