incubator-cassandra-user mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From aaron morton <aa...@thelastpickle.com>
Subject Re: Configuring the keyspace correctly - NTS
Date Wed, 14 Sep 2011 22:15:13 GMT
The strategy_options for NTS accept the data centre name and the rf, [{<dc_name> : <dc_rf>}]

Where the DC name comes from the snitch, so…

SimpleSnitch (gotta love this guy, in there day in day out putting in the hard yards) puts
all the nodes in "datacenter1" which is why thats in the defaults. 

RackInferringSnitch (or the "Hollywood Snitch" as I call it) puts the them in a DC named after
the second octet of the IP. So 130 in your  case. 

PropertyFileSnitch does whats in the cassandra-topology.properties file. EC2Snitch uses the
EC2 Region. Brisk snitch does it's thing.

If you want to use 130 you should be using the RackInferringSnitch, if you want to use human
names use either the SimpleSnitch or the PropertyFileSnitch. Property File Snitch has a default
catch all DC, see the cassandra-topology.properties file. 

Cheers

-----------------
Aaron Morton
Freelance Cassandra Developer
@aaronmorton
http://www.thelastpickle.com

On 15/09/2011, at 9:43 AM, Anthony Ikeda wrote:

> Okay, in a previous post, it was stated that I could use a NetworkTopologyStrategy in
a singel data centre by setting up my keyspace with:
> 
> create keyspace KeyspaceDEV
> 
>     with placement_strategy = 'org.apache.cassandra.locator.NetworkTopologyStrategy'
> 
>     and strategy_options=[{datacenter1:3}];
> 
>  
> Whereby my understanding is that:
> 
> [{datacenter1:3}]
> 
> represents:
> 
> 
> 1 Datacentre
> 3 nodes in that datacentre
> My infrastructure team were recommended to instead of use "datacenter1" to use the second
value in the IP address:
> x.130.x.x
> 
> [{130:3}]
> 
> However, when trying to access the keyspace the following error was return:
> "May not be enough replicas present to handle consistency level"
> 
> When I rebuilt the keyspace using the "datacenter1" semantic, it worked fine.
> 
> My guess is that there is some correlation between the "130" value and either the rpc_address
or listen_address. Am I correct in thinking this?
> 
> I don't have access to the se configurations so I'm just going out on a whim here trying
to figure out why using the "130" form the IP address would cause the error.
> 
> Anthony
> 
> 
> 


Mime
View raw message