[
https://issues.apache.org/jira/browse/CASSANDRA-18398?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17717317#comment-17717317
]
David Capwell commented on CASSANDRA-18398:
-------------------------------------------
bq. the result is that we may miss data
This would be a data loss issue for users, we have the data in Data.db, but we
fail to return it due to such an issue, so still data loss.
bq. It is very difficult to imagine a corruption of both the partition and row
index in a way that successfully leads to the wrong key being accepted.
I guess it all depends on how defensive we are... given this is prefix based, I
assume we need to validate the key once the index returns offsets where as we
don't need to in the current "big" format (as its full), so with BTI we are
less likely than BIG to return the wrong partition, but both suffer from
failure to return a partition due to this type of corruption.
> CEP-25: Trie-indexed SSTable format
> -----------------------------------
>
> Key: CASSANDRA-18398
> URL: https://issues.apache.org/jira/browse/CASSANDRA-18398
> Project: Cassandra
> Issue Type: Improvement
> Components: Local/SSTable
> Reporter: Branimir Lambov
> Assignee: Branimir Lambov
> Priority: Normal
> Fix For: 5.x
>
> Time Spent: 17h
> Remaining Estimate: 0h
>
> Implementation of Big Trie-Indexed (BTI) SSTable format, per
> [CEP-25|https://cwiki.apache.org/confluence/display/CASSANDRA/CEP-25%3A+Trie-indexed+SSTable+format].
--
This message was sent by Atlassian Jira
(v8.20.10#820010)
---------------------------------------------------------------------
To unsubscribe, e-mail: commits-unsubscr...@cassandra.apache.org
For additional commands, e-mail: commits-h...@cassandra.apache.org
