incubator-cassandra-user mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Matthew Dennis <mden...@datastax.com>
Subject Re: Internal error processing insert
Date Tue, 15 Feb 2011 01:08:38 GMT
I had actually meant to (and thought I did) type "greater than zero and less
than *or equal* to number of nodes".

That being said, you usually do want it less than the number of nodes in the
cluster because otherwise your cluster essentially has the same performance
as a single node.  In general (fuzzy) performance is num_nodes / RF.  If you
have 9 nodes and RF=3 you have performance roughly along the lines of having
3 boxen.  If you have N nodes and RF=N, you have roughly the same
performance as... one machine.

What Arron says about RF and Consistency Level (CL) is also correct.  RF=2
implies CL.Q requires both nodes meaning you can't lose a single node and
still read/write at CL.Q

Just to be pedantic though, you have have a cluster of N nodes (for large N)
and read/write at CL.Q even after losing some nodes.  For example, if you
have 10 nodes and RF=10 you only need 6 nodes up to read/write at CL.Q

On Mon, Feb 14, 2011 at 6:58 PM, Aaron Morton <aaron@thelastpickle.com>wrote:

> He probably meant in production. When playing around, and if you only have
> 2 nodes, you can set it to 2.
>
> From memory RF of 2 means the Quorum is also 2, so you cannot afford to
> lose one. Thats fine for playing.
>
> Aaron
>
>
> On 15 Feb, 2011,at 01:51 PM, mcasandra <mohitanchlia@gmail.com> wrote:
>
>
>
> mcasandra wrote:
> >
> > In earlier post same thread you mentioned that replication factor should
> > be set to less than N.
> >
> > Currently I am testing on 2 node cluster and I was able to set
> > replication_factor to 2 (=N) and also when I did cfstats (I don't quite
> > understand cfstats in detail) and see some activity on both nodes now
> >
> > [default@twissandra] update keyspace twissandra with
> replication_factor=2;
> > 3ed6b708-389b-11e0-993f-b7fa7ed61af9
> > [default@twissandra] set users['b']['add']='111';
> > Value inserted.
> >
> > On the other node when I ran nodetool with -h as localhost, I see:
> >
> >
> > Column Family: users
> > SSTable count: 0
> > Space used (live): 0
> > Space used (total): 0
> > Memtable Columns Count: 2
> > Memtable Data Size: 52
> > Memtable Switch Count: 0
> > Read Count: 3
> > Read Latency: NaN ms.
> > Write Count: 2
> > Write Latency: NaN ms.
> > Pending Tasks: 0
> > Key cache capacity: 200000
> > Key cache size: 0
> > Key cache hit rate: NaN
> > Row cache: disabled
> > Compacted row minimum size: 0
> > Compacted row maximum size: 0
> > Compacted row mean size: 0
> >
> >
> Sorry I meant to say someone (Matthew Dennis) on the thread mentioned that
> replication factor should be less than no. of nodes.
> --
> View this message in context:
> http://cassandra-user-incubator-apache-org.3065146.n2.nabble.com/Internal-error-processing-insert-tp6025740p6025882.html
> Sent from the cassandra-user@incubator.apache.org mailing list archive at
> Nabble.com.
>
>

Mime
View raw message