[
https://issues.apache.org/jira/browse/CASSANDRA-193?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=12759994#action_12759994
]
Stu Hood commented on CASSANDRA-193:
------------------------------------
More thoughts (in no particular order, and probably disagreeing with previous
decisions):
1. Since Memtables are always flushed after a bounded period (I think?) it is
possible to completely ignore them.
2. Due to 1, invalidating a range for every Memtable change is dumb/overkill.
It would be much more efficient to only invalidate during minor compactions.
3. Again, due to 1, we only care about the data affected during a major
compaction, so we can validate the tree during every major compaction.
If we hook tree validation into major compactions:
* Handling TreeRequests/TreeResponses would be queued until the next major
compaction,
* The compactor would acquire a lock on the tree for the CF, validate it, and
then handle outstanding TreeRequests/TreeResponses.
> Proactive repair
> ----------------
>
> Key: CASSANDRA-193
> URL: https://issues.apache.org/jira/browse/CASSANDRA-193
> Project: Cassandra
> Issue Type: New Feature
> Components: Core
> Reporter: Jonathan Ellis
> Assignee: Stu Hood
> Fix For: 0.5
>
> Attachments: CASSANDRA-193.diff
>
>
> Currently cassandra supports "read repair," i.e., lazy repair when a read is
> done. This is better than nothing but is not sufficient for some cases (e.g.
> catastrophic node failure where you need to rebuild all of a node's data on a
> new machine).
> Dynamo uses merkle trees here. This is harder for Cassandra given the CF
> data model but I suppose we could just hash the serialized CF value.
--
This message is automatically generated by JIRA.
-
You can reply to this email to add a comment to the issue online.
