Hi again :-)
O.k... New problem...
I have an Amazon EC2 node with 4 "CPUs" and 7.5 GB of RAM.
Running CommitLog on 1 disk and data on another.
Cassandra 0.4.0 - (yes I have checked... correct version :-P)
6GB set in the cassandra.in.sh.
I started throwing data at it, without problems.
All of a s
A few things to try:
1. Enable verbose GC logging to see if your JVM is dying under GC load.
2. pkill -3 java will dump some nice stack traces from all running
threads, could be some clues there.
Dan Larsen wrote:
> Hi again :-)
>
> O.k... New problem...
> I have an Amazon EC2 node with 4 "CPUs"
Thanks for the tips Eric.
I was just about to try it, when I noticed, it had become responsive
again.
It took exactly 1 hour, before it was done!...
But when I restart now, it's ready almost immediatly... Weird stuff!!
I will try out your tips, next time this happens!
It sounds like, it's p
if you swamp it with inserts faster than it can write them, it will
start spending more and more time trying to GC. that's what's
happening... trunk is smarter about this and will stop accepting
writes before it gets to that point, but for 0.4 you just need to be a
little careful.
-Jonathan
On
Hi Jonathan
I could reproduce this on the Super1 super column family that comes
pre-packaged in cassandra's config file.
Here is the small program I wrote:
public class TestCassandra {
public static void main(String[] args)
{
try {
TestCassandra cassandra = new TestC
Thanks, I'll have a look.
On Fri, Oct 9, 2009 at 11:58 AM, Ramzi Rabah wrote:
> Hi Jonathan
>
> I could reproduce this on the Super1 super column family that comes
> pre-packaged in cassandra's config file.
> Here is the small program I wrote:
Just a note to inform the list of a python module for cassandra:
http://bitbucket.org/_ben_/cassdict/overview/
For better or worse Cassdict attempts to emulate a standard python
dictionary as an interface to the python thrift api.
It simplifies thing as far as it goes. Inevitably there's no
One suggestion -- consider subclassing DictMixin, instead of dict
itself. DictMixin is designed for creating things that have a
dict-like api, but aren't actually dicts, whereas dict itself
obviously isn't. :) Tends to work out cleaner in my experience.
Just my two cents,
-Jonathan
On Fri, Oct
That is exactly, what I have discovered by now ;-)
Looking forward for the next release then!
Thanks Jonathan!
Best
Dan
On 09/10/2009, at 16.21, Jonathan Ellis wrote:
if you swamp it with inserts faster than it can write them, it will
start spending more and more time trying to GC. that's w
Hi Tom,
We're using an internal maven repository to store both the Java thrift libs
and the Java cassandra client. As long as your project doesn't depend on
dynamically generated thrift objects (for purposes outside of cassandra)
that will work fine.
Cheers,
Joel
On Wed, Oct 7, 2009 at 3:09 PM, T
Oops, I should have replied to Todd, not Tom: we're doing what Tom described
and it works fine for us.
On Fri, Oct 9, 2009 at 12:02 PM, Joel Meyer wrote:
> Hi Tom,
> We're using an internal maven repository to store both the Java thrift libs
> and the Java cassandra client. As long as your proje
Cheers, I looked at dictmixin, thought may as well subclass dict,
but i'll check it out again now, thanks for the tip.
Ben
On 9 Oct 2009, at 18:25, Jonathan Ellis wrote:
One suggestion -- consider subclassing DictMixin, instead of dict
itself. DictMixin is designed for creating things th
I've got a guy doing a code test for us and he has some questions
about custom partitioners:
http://gist.github.com/205537
Wondering if anyone could chime in.
Thanks!
--Joe
The patch attached to
https://issues.apache.org/jira/browse/CASSANDRA-484 should fix this
for you. Thanks for the test case!
thanks a lot for the fix :)
On Fri, Oct 9, 2009 at 6:46 PM, Jonathan Ellis wrote:
> The patch attached to
> https://issues.apache.org/jira/browse/CASSANDRA-484 should fix this
> for you. Thanks for the test case!
>
15 matches
Mail list logo