[
https://issues.apache.org/jira/browse/CASSANDRA-3648?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13173001#comment-13173001
]
Sylvain Lebresne commented on CASSANDRA-3648:
---------------------------------------------
I think that for block checksums we're always fine, at least currently, since
even for streaming of compressed file, we first decompress them before
streaming and thus validate the checksum at the same time.
For full file hashes, as you said we won't add a check on the read path. So
adding checksum validation before repair doesn't provide bulletproof protection
against propagation of corruption, but for the case where the corruption makes
the file unreadable, read repair would not propagate those type of corruption
(since the read would fail), but streaming would (because it just take two
offset based on the index file and stream between those two offsets). So at
least we could avoid that kind of corruption propagation, even though that is
not the worst case for corruption.
> Repair should validate checksums before streaming
> -------------------------------------------------
>
> Key: CASSANDRA-3648
> URL: https://issues.apache.org/jira/browse/CASSANDRA-3648
> Project: Cassandra
> Issue Type: New Feature
> Reporter: Jonathan Ellis
> Priority: Minor
> Fix For: 1.2
>
>
> We have block checksums and sha1 full-file hashes, but currently we do not
> validate these before streaming in a repair.
--
This message is automatically generated by JIRA.
If you think it was sent incorrectly, please contact your JIRA administrators:
https://issues.apache.org/jira/secure/ContactAdministrators!default.jspa
For more information on JIRA, see: http://www.atlassian.com/software/jira
