[
https://issues.apache.org/jira/browse/CASSANDRA-4047?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13817758#comment-13817758
]
Brandon Williams edited comment on CASSANDRA-4047 at 11/8/13 10:28 PM:
-----------------------------------------------------------------------
Attaching what I had here, mostly-rebased against 2.0, before I hit a snag. It
doesn't quite compile against 2.0 because now streaming is totally different,
but on 1.2 it would work if you manually inserted the bulk hints with cqlsh.
The snag I hit was actually inserting the hints from the bulk loader. While
it's fairly simple to explain, in that you just need to insert a hint for a
failure on any replica that does succeed, I couldn't get the information I
needed (ks/cf name for the insert, and range from the filename) out of the
streaming callback at the time.
Can you take over, Yuki?
was (Author: brandon.williams):
Attaching what I had here, mostly-rebased against 2.0, before I hit a snag. It
doesn't quite compile against 2.0 because now streaming is totally different,
but on 1.2 it would work if you manually inserted the bulk hints with cqlsh.
The snag I hit was actually inserting the hints from the bulk loader. While
it's fairly simple to explain, in that you just need to insert a hint for a
failure on any replica that does succeed, I couldn't get the information I
needed (ks/cf name for the insert, and range from the filename) out of the
streaming callback at the time.
Can you over, Yuki?
> Bulk hinting
> ------------
>
> Key: CASSANDRA-4047
> URL: https://issues.apache.org/jira/browse/CASSANDRA-4047
> Project: Cassandra
> Issue Type: Improvement
> Components: Core
> Reporter: Brandon Williams
> Assignee: Yuki Morishita
> Fix For: 2.0.3
>
> Attachments: 4047-wip.txt
>
>
> With the introduction of the BulkOutputFormat, there may be cases where
> someone would like to tolerate node failures and have the job complete, but
> afterwards since we streamed they have to repair or rely on read repair. We
> don't currently have any way of hinting streams, but a node could take a
> snapshot before acknowledging the stream session, then remember to send the
> files in the snapshot to the unavailable nodes when they come back up. This
> isn't quite ideal since of course the node may have compacted these files,
> however it's much simpler than any sort of key tracking at this scale.
--
This message was sent by Atlassian JIRA
(v6.1#6144)
