MySQL Cluster (NDB) spread data (rows) to 4 data-nodes servers
Posted by:
Ali Saeidi
Date: December 02, 2023 08:12AM
I want to spread my MySQL rows to 4 servers (Horizontal sharding). I searched, and I found that I have to use NDB cluster (MySQL cluster). I made a MySQL server/client with NDB manager server and 4 data nodes servers and this is my manager configuration :
[ndbd default]
NoOfReplicas=1
[ndb_mgmd]
NodeId=1
hostname=192.168.159.133
datadir=/var/lib/mysql-cluster
[ndbd]
NodeId=2
hostname=192.168.159.131
datadir=/home/db
[ndbd]
NodeId=3
hostname=192.168.159.132
datadir=/home/db
[ndbd]
NodeId=4
hostname=192.168.159.137
datadir=/home/db
[mysqld]
NodeId=20
hostname=192.168.159.133
I know replication means make a backup of each row to another server and I don't want it (I just want to spread my data (rows) to 4 servers, not making backup) so I set NoOfReplicas to 1 to have 4 nodegroups ...
Am I right ? Does my data (rows) spread on that 4 data-node servers ?
If I right, what is different between doing this, or just using Innodb and make foreign partitions (on remote servers) ? Because it's a lot easier to make partition (key) by my own and attach each partition to remote servers !!!
Is using NDB has more advantage (about what I want to do (spread data, not replication (making backup))) ?
And final question, is it right to have NDBCLUSTER table with also own partitions ???!!! What is that mean, anyway ? For example, I have 4 data-nodes and for a NDBCLUSTER table, I create 8 partitions too (key partition) ! How that's work? Does it create 8 partitions on each data-node ? Or just split those partitions on those 4 data-nodes server (for example, each data-node holds 2 partitions) ?
Subject
Views
Written By
Posted
MySQL Cluster (NDB) spread data (rows) to 4 data-nodes servers
353
December 02, 2023 08:12AM
223
December 05, 2023 01:19PM
Sorry, you can't reply to this topic. It has been closed.
Content reproduced on this site is the property of the respective copyright holders.
It is not reviewed in advance by Oracle and does not necessarily represent the opinion
of Oracle or any other party.