incubator-cassandra-user mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Matthew Allen <matthew.j.al...@gmail.com>
Subject Changing replication strategies
Date Mon, 07 Apr 2014 03:12:28 GMT
Hi all.

To move to a multi-dc setup, I am in the process of preparing our
independent production DC's.

Part of this means moving from SimpleStragtegy (RF=5) to
NetworkTopologyStrategy, but I'm still a bit confused as to some core
aspects of replication.

The setup will be basic,
 - DC_NSW will contain 9 nodes in a single rack.
 - DC_VIC will contain 9 nodes in a single rack
 - Initially the Replication Factor will be DC_NSW:5.
 - Once DC_VIC comes online, the Recplication Factor will be DC_NSW:5,
DC_VIC:5

My understanding is that SimpleStrategy just 'walks the token ranges' to
identify where to place replicas, whereas NetworkTopology will try to place
the data in differing racks within the same datacenter (which will not be
applicable to us given we have 1 rack per datacenter)

1. When does Cassandra walk the tokens/nodes to idenfity which hosts should
contain replicas.  At every insert, Keyspace creation, Keyspace
modification or ... ?
2. Is there a command or JMX entry that displays which ranges a server is
repsonsible for (not its primary range, but ranges it is acting as a
replica for) ?
3. Hypothetically, if we defined multiple racks per datacenter, and updated
from Simple to Network
 - Then the replicas for a particular range may change ?
 - If so, how would Cassandra deal with this in a live environment ?

Thanks for your help

Matt

Mime
View raw message