Mailing-List: contact cassandra-commits-help@incubator.apache.org;
 run by ezmlm
Precedence: bulk
Reply-To: cassandra-dev@incubator.apache.org
Message-ID: <1025327505.225961263424854427.JavaMail.jira@brutus.apache.org>
Date: Wed, 13 Jan 2010 23:20:54 +0000 (UTC)
From: "Gary Dusbabek (JIRA)" <jira@apache.org>
To: cassandra-commits@incubator.apache.org
Subject: [jira] Commented: (CASSANDRA-696) Bootstrapping doesn't work on new
 clusters
In-Reply-To: <855283255.224341263421194494.JavaMail.jira@brutus.apache.org>
MIME-Version: 1.0
Content-Type: text/plain; charset=utf-8
Content-Transfer-Encoding: 7bit


    [ https://issues.apache.org/jira/browse/CASSANDRA-696?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=12800005#action_12800005 ] 

Gary Dusbabek commented on CASSANDRA-696:
-----------------------------------------

To be precise, it looks like anything still in the memtables will not be streamed to the bootstrapping node since we're not blocking on the table.flush() call.  I haven't researched it enough, but I suspect the same thing will happen in an established cluster: everything already committed to SSTables gets streamed, and anything still left in the memtable is left behind.  It's just that in a brand new cluster, there never are any SSTables in the first place--it makes the bug more obvious.

> Bootstrapping doesn't work on new clusters
> ------------------------------------------
>
>                 Key: CASSANDRA-696
>                 URL: https://issues.apache.org/jira/browse/CASSANDRA-696
>             Project: Cassandra
>          Issue Type: Bug
>          Components: Core
>    Affects Versions: 0.5
>            Reporter: Gary Dusbabek
>            Assignee: Gary Dusbabek
>            Priority: Minor
>         Attachments: wait_for_memtable_flush.patch
>
>
> This is an edge case.
> 1. start a clean 3 node cluster with autobootstrap on.
> 2. load some data.
> 3. bootstrap in a 4th node.
> the logs in the 4th node will indicate that data was not received.  If you restart the cluster in between steps 1 and 2, or 2 and 3, boot strapping works fine.  
> I find that waiting on the table flush when making the streaming request solves the problem (see patch).

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