[
https://issues.apache.org/jira/browse/CASSANDRA-193?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=12759990#action_12759990
]
Stu Hood edited comment on CASSANDRA-193 at 9/26/09 11:00 PM:
--------------------------------------------------------------
I've been working on this ticket a bit more in the past few days:
* Added o.a.c.service.AntiEntropyService - Maintains trees for each CF, and
accepts invalidations when values change.
Still TODO:
* Implement TreeRequestVerbHandler/TreeResponseVerbHandler - The AEService on
a first endpoint will periodically wake up and send a TreeRequest to a replica.
The replica endpoint will handle the TreeRequest by validating one or all of
its MerkleTrees, and responding with a TreeResponse. Handling the TreeResponse
on the first endpoint will involve validating the local tree, and then
comparing the two trees.
* Validation is the only part that is fuzzy here: we need to iterate over
keys in each CF (essentially, a major compaction, except that we can skip
processing for anything that is still valid in the tree).
* Begin implementing the actual repair step - There isn't a design for this
part yet: any thoughts would be appreciated. The output of the
TreeRequest/TreeResponse conversation will be a list of ranges in a given CF
that disagree between the two endpoints.
EDIT: The code is still located at:
http://github.com/stuhood/cassandra-anti-entropy/
was (Author: stuhood):
I've been working on this ticket a bit more in the past few days:
* Added o.a.c.service.AntiEntropyService - Maintains trees for each CF, and
accepts invalidations when values change.
Still TODO:
* Implement TreeRequestVerbHandler/TreeResponseVerbHandler - The AEService on
a first endpoint will periodically wake up and send a TreeRequest to a replica.
The replica endpoint will handle the TreeRequest by validating one or all of
its MerkleTrees, and responding with a TreeResponse. Handling the TreeResponse
on the first endpoint will involve validating the local tree, and then
comparing the two trees.
* Validation is the only part that is fuzzy here: we need to iterate over
keys in each CF (essentially, a major compaction, except that we can skip
processing for anything that is still valid in the tree).
* Begin implementing the actual repair step - There isn't a design for this
part yet: any thoughts would be appreciated. The output of the
TreeRequest/TreeResponse conversation will be a list of ranges in a given CF
that disagree between the two endpoints.
> 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.
