cassandra-commits mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From "Jaakko Laine (JIRA)" <>
Subject [jira] Commented: (CASSANDRA-558) optimize local writes
Date Tue, 17 Nov 2009 14:31:39 GMT


Jaakko Laine commented on CASSANDRA-558:

1) I was thinking of having some helper function doTheStuff(WriteResponseHandler responseHandler)
and then set the responsehandler if it was not null. Don't know if this would be enough, though.

3) It can be local if cluster size is only slightly bigger than replication factor: (a) Suppose
we have a cluster of four nodes (1, 2, 3 and 4) with replication factor 3. (b) Suppose that
write is done, for which 1, 2 and 3 are responsible for storing replicas. (c) Suppose that
this write is directed to node 4, and nodes 2 and 3 are down at this time. In this case one
write will go to node 1, and node 4 will store hinted replicas for 2 and 3. Even if only one
of these nodes (1, 2 or 3) is down, the hinted replica will go to the local node to which
write operation was directed. This is because getHintedMapForEndpoints fills the list of endpoints
with local endpoint in case it cannot get enough of unused live endpoints. But as mentioned,
this is a very rare case, so probably not worth the trouble.

5) I'll test this more in the morning.

> optimize local writes
> ---------------------
>                 Key: CASSANDRA-558
>                 URL:
>             Project: Cassandra
>          Issue Type: Improvement
>            Reporter: Jonathan Ellis
>            Assignee: Jonathan Ellis
>            Priority: Minor
>             Fix For: 0.5
>         Attachments: 0001-CASSANDRA-558-make-local-insert-skip-MessagingServic.txt, 0002-add-WriteResponseHandler-combining-the-important-parts.txt,
0003-rename.txt, 0004-two-trivial-changes.patch

This message is automatically generated by JIRA.
You can reply to this email to add a comment to the issue online.

View raw message