You are viewing a plain text version of this content. The canonical link for it is here.
Posted to user@cassandra.apache.org by Michael Greene <mi...@gmail.com> on 2009/09/29 00:27:35 UTC
Cassandra cluster setup [WAS Re: usage]
This is a new thread continued from the Facebook-usage thread.
Cassandra automatically shards your data based on the Partitioner you
have setup in storage-conf.xml. The copies are controlled by the
ReplicationFactor setting in the same configuration file. If all your
nodes are in the same data center, then the default
ReplicaPlacementStrategy of RackUnawareStrategy should be fine.
Michael
On Mon, Sep 28, 2009 at 5:17 PM, Kirill A. Korinskiy
<ca...@catap.ru> wrote:
> But how in cassandra now make a sharding by nodes?
>
> For example: I have a six nodes and I need make a 2 copy of data and
> sharding data by all cluster. For example all nodes of my cluster stay
> on one dc.
>
> How do it?
>
> --
> wbr, Kirill
>
Re: Cassandra cluster setup [WAS Re: usage]
Posted by Jonathan Ellis <jb...@gmail.com>.
On Mon, Sep 28, 2009 at 5:34 PM, Kirill A. Korinskiy
<ca...@catap.ru> wrote:
>> Cassandra automatically shards your data based on the Partitioner you
>> have setup in storage-conf.xml. The copies are controlled by the
>> ReplicationFactor setting in the same configuration file. If all your
>> nodes are in the same data center, then the default
>> ReplicaPlacementStrategy of RackUnawareStrategy should be fine.
>>
>
> ok. But on my test all my data go to two nodes.
Until load balancing (ticket 192) is done, you need to choose the node
tokens so that the keys are spread among them evenly. You can do this
either with InitialToken in the configuration, or by using the token
updater script in bin/.
-Jonathan
Re: Cassandra cluster setup [WAS Re: usage]
Posted by "Kirill A. Korinskiy" <ca...@catap.ru>.
At Mon, 28 Sep 2009 18:07:03 -0500,
Michael Greene <mi...@gmail.com> wrote:
>
> Which partitioner are you using?
org.apache.cassandra.dht.RandomPartitioner
> What are the tokens for the nodes or what does bin/nodeprobe ring
> output?
Starting Token Ending Token Size Address Ring
155352990208345788173011248201827004586 20283731607556365006538603533864462732 2 67.218.100.94 |<--|
20283731607556365006538603533864462732 21836380263552532517023620922568993549 2 67.218.100.93 | ^
21836380263552532517023620922568993549 27825817942936339100357854565774016889 2 67.218.100.92 v |
27825817942936339100357854565774016889 56705206375993081178630024509360001548 2 67.218.100.116 | ^
56705206375993081178630024509360001548 154786577292345968165033245777473466277 2 67.218.100.115 v |
154786577292345968165033245777473466277 155352990208345788173011248201827004586 2 67.218.100.117 |-->|
> Are you setting the initial token or allowing Cassandra to
> choose one?
No, initialtoken is empty and Cassandra chose itinital toke.
> What do your keys look like? Are they well-distributed?
>
So. I try to insert as key md5(X) where X is sequences 1..100000 and
all key seting on 67.218.100.116 and 67.218.100.115.
> Michael
>
> On Mon, Sep 28, 2009 at 5:34 PM, Kirill A. Korinskiy
> <ca...@catap.ru> wrote:
> > At Mon, 28 Sep 2009 17:27:35 -0500,
> > Michael Greene <mi...@gmail.com> wrote:
> >>
> >> This is a new thread continued from the Facebook-usage thread.
> >>
> >
> > sure
> >
> >> Cassandra automatically shards your data based on the Partitioner you
> >> have setup in storage-conf.xml. The copies are controlled by the
> >> ReplicationFactor setting in the same configuration file. If all your
> >> nodes are in the same data center, then the default
> >> ReplicaPlacementStrategy of RackUnawareStrategy should be fine.
> >>
> >
> > ok. But on my test all my data go to two nodes.
> >
> > --
> > wbr, Kirill
> >
--
wbr, Kirill
Re: Cassandra cluster setup [WAS Re: usage]
Posted by Michael Greene <mi...@gmail.com>.
Which partitioner are you using?
What are the tokens for the nodes or what does bin/nodeprobe ring
output? Are you setting the initial token or allowing Cassandra to
choose one?
What do your keys look like? Are they well-distributed?
Michael
On Mon, Sep 28, 2009 at 5:34 PM, Kirill A. Korinskiy
<ca...@catap.ru> wrote:
> At Mon, 28 Sep 2009 17:27:35 -0500,
> Michael Greene <mi...@gmail.com> wrote:
>>
>> This is a new thread continued from the Facebook-usage thread.
>>
>
> sure
>
>> Cassandra automatically shards your data based on the Partitioner you
>> have setup in storage-conf.xml. The copies are controlled by the
>> ReplicationFactor setting in the same configuration file. If all your
>> nodes are in the same data center, then the default
>> ReplicaPlacementStrategy of RackUnawareStrategy should be fine.
>>
>
> ok. But on my test all my data go to two nodes.
>
> --
> wbr, Kirill
>
Re: Cassandra cluster setup [WAS Re: usage]
Posted by "Kirill A. Korinskiy" <ca...@catap.ru>.
At Mon, 28 Sep 2009 17:27:35 -0500,
Michael Greene <mi...@gmail.com> wrote:
>
> This is a new thread continued from the Facebook-usage thread.
>
sure
> Cassandra automatically shards your data based on the Partitioner you
> have setup in storage-conf.xml. The copies are controlled by the
> ReplicationFactor setting in the same configuration file. If all your
> nodes are in the same data center, then the default
> ReplicaPlacementStrategy of RackUnawareStrategy should be fine.
>
ok. But on my test all my data go to two nodes.
--
wbr, Kirill