Production configuration for Cluster mode

22 views
Skip to first unread message

Avise Sudhakar Rao

unread,
Oct 29, 2012, 1:09:58 PM10/29/12
to mongod...@googlegroups.com
Hello Team,

Please suggest me what should be the production configuration for Sharding/Replica Set architecture.

Currently database size is of 34GB in Replica Set configuration with 4,00,000 lakhs writes per sec. checked by iostat and soon expecting 20,00,000
lakhs writes in a 1 month gap.

Due to that planning for Sharding with below configurations, please correct for any changes:-

a.       Shard Servers

                                                                                                   i.      Configuration

1.       RAM: 38 GB

2.       Hard Disk: 350GB

3.       Maximize the number of disks and RAID controller cache and do Striping, typically 1MB stripes. No mirroring.

4.       4 CPU’s

 

b.      Arbiters

                                                                                                   i.      Configuration

1.       RAM: 4GB

2.       Hard Disk:100GB

c.       MongoDBConfig Servers

                                                                                                   i.      Configuration

1.       RAM: 8GB

2.       Hard Disk: 150GB

d.      Mongos (ConfigDB):

                                                                                                   i.      Configuration

1.       RAM: 4GB

2.       Hard Disk: 100G


AnAnd also suggest me how much oplogsize and chunksize should I need to place in each shard servers? And also let me know what all the important settings should I enable in the mongodb.conf file


Th  Thanks,

      Sudhakar

Sudhakar

unread,
Nov 5, 2012, 8:08:12 AM11/5/12
to mongod...@googlegroups.com
Hi Team,

Can I have any suggestions on oplogsize on the shardservers and chunksize on the mongos server for the production setup?

Thanks,
Sudhakar

David Hows

unread,
Nov 23, 2012, 1:07:22 AM11/23/12
to mongod...@googlegroups.com
Hi Sudhakar,

The oplog size and chunksize will depend on the size of your documents.

The oplog size will also depend on how much replication delay you wish to allow for within your systems until they go stale and need to be manually resynchronised.

The chunk size will depend (basically) on how regularly you wish to be migrating chunks and how much difference in data volume between shards.

Cheers,

David



Reply all
Reply to author
Forward
0 new messages