bulk load glitch

[Brian Jeltema <br...@digitalenvoy.net>]

I'm attempting to perform a bulk load by calling the jmx:bulkLoad method on several nodes in parallel. In a Casssandra log
file I see a few occurrences of the following:

 INFO [GossipTasks:1] 2012-07-02 10:12:33,626 Gossiper.java (line 759) InetAddress /10.4.0.3 is now dead.
ERROR [GossipTasks:1] 2012-07-02 10:12:53,630 StreamOutSession.java (line 232) StreamOutSession /10.4.0.3 failed because {} died or was restarted/removed
 INFO [GossipStage:1] 2012-07-02 10:18:40,008 Gossiper.java (line 745) InetAddress /10.4.0.3 is now UP

I can't tell whether the bulk load process recovered from the transient dead node, or whether I need to start over.

Does anybody know?

Brian

This is Cassanrda 1.0.5, by the way.

Re: bulk load glitch

[Brandon Williams <dr...@gmail.com>]

On Mon, Jul 2, 2012 at 10:35 AM, Brian Jeltema
<br...@digitalenvoy.net> wrote:
> I can't tell whether the bulk load process recovered from the transient dead node, or whether I need to start over.
>
> Does anybody know?

You need to start over if the failure detector tripped, but it will
retry a few times for regular network errors.

-Brandon