Using multiple disks with Riak

Karsten Thygesen karthy at netic.dk
Wed Jan 5 08:23:54 EST 2011


Hi Yusuf

On Jan 5, 2011, at 13:28 , yusuf şahin wrote:

> Hello Riak Users!
> 
> 
> I am working on a testing Riak cluster and have some questions about disks on nodes.
> 
> 1 - How big or small disk arrays we should use with riak nodes when we have 1GBS network between nodes and have 8GB ram on each nodes. Does nodes have limitations like ram cpu, bandwith while storing too big data consist of many small files. 

Many small files will require many keys and keys are in memory when using Bitcask as backend. So memory will probably be your limitation.

> 
> 2 - Is it ok to use different size of disks with riak nodes. For exaple, 100GB for 1st node , 300GB for second node, 500GB for third node etc. How will Riak handle the data distrubition and replication in this situation in a large cluster of nodes.  (if we have 20 or 50 nodes in the cluster)

Yes, it is OK, but Riak does not take available diskspace into consideration when distributing data. All nodes gets a fair chunk of vnodes and thereby the approximate same amount of data to handle. So capacity wise, you are limited by the smallest server.

> 
> 3- When disk on one of nodes become full, what will happen ?

> 
>     Node will continue serving files but it will not accept any new data, Node will accept any read, delete request but will not accept any write request so new files will be written to another node with enough disk space ?
>     Node will stop serving files and will go dead. So the files in that node will replicated to another node ?
>    We can mount new disks to that node to increase total disk volume ?    
>    Which condition is ture ?

The node will die and others will take over.

> 
> 4- Is Riak support multiple disks for data storage, if yes have can we make this?

Bitcask is a single directory of directories of files - the easiest way to use multiple disks is to use the native raid feature of your operatingsystem. We use Solaris and ZFS to handle raid among multiple disks - this also gives us the opportunity to make snapshots for backup purposes and we can extend the filesystem on the run...

Best regards,
Karsten
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.basho.com/pipermail/riak-users_lists.basho.com/attachments/20110105/0d24457a/attachment-0001.html>


More information about the riak-users mailing list