cassandra-user mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Vijay <>
Subject Re: Configuring Cassandra
Date Tue, 08 Dec 2009 19:24:30 GMT
a) Multiple DCs: Can I control that each key should store in each and every
data center? If it is possible, an example configuration will be helpful.
    you cannot.... and if you did the ring will break... consider Cassandra
in multiple DC as one big Ring..... We use Cassandra to have copies of all
data in all the DC's


On Tue, Dec 8, 2009 at 5:56 AM, Mark Robson <> wrote:

> I'm not an expert in Cassandra yet but I do know a little, here are some
> (attempted) answers:
> 2009/12/8 Rakesh Sharma <>
> b) Apart from nodeprobe and Jconsole is there any other node management
>> tool?
> You can use any JMX application, apparently, to monitor it. This could be
> done e.g. via a command line script (assuming there is a convenient JMX CLI)
> and then hooked into your monitoring. I'd look there first. I have not tried
> this (yet).
> d) Can I control where a node will be added in the ring? How is it decided?
>> Primarily, I want to understand how the hot-spots can be handled?
> You can pick the token manually before first startup; it gets read from the
> config file.
> Detecting hot-spots is not easy. If you use the OrderedPartitioner, then
> hot spots tend to occur where you put all your keys. The fix is to spread
> out your keys, or bunch up your nodes, but if you use time-based keys or
> similar, your key distribution will change over time so you can't win that
> way.
> Nodes can be decommissioned now (as of 0.5 I think) with the nodeprobe
> decommission command. You can then stop the node, clear all its data and
> re-bootstrap it with a different token. This is not an automatic or
> particularly fast or convenient process.
>> e) How a node data backup should be taken such that reads and writes at
>> that time are not effected (read or write locks)? Any tool/utility that can
>> be used?
> The nodeprobe snapshot can take a snapshot of an individual node. You can
> of course schedule these to happen at (about) the same time.
>> f) Can I take backups from each node and merge them on a single node? Or
>> it's not at all required?
> Dunno, sounds nontrivial.
> Mark

View raw message