Here is a working configuration.

[Joseph Kouyoumjian]

Hello, all.

Here is a configuration I got to work with MySQL Max 5.0.18.

I have two management servers named "emerald" and "ruby".
I have two data nodes named "quartz" and "amber".
I have two SQL nodes named "jade" and "coral".


Here is the /var/lib/mysql-cluster/config.ini file for emerald and ruby:

(Note that even though quartz and amber are data nodes, the management nodes are configured to allow connections from the API for the purpose of restoring backups with ndb_restore. Note also that the IDs are out of order becuase of the way I added machines. I had to assign IDs explicitly when setting up two management nodes).

[NDBD DEFAULT]
NoOfReplicas=2
DataMemory=200M
IndexMemory=100M

[NDB_MGMD]
id=1
hostname=emerald
datadir=/var/lib/mysql-cluster

[NDBD]
id=2
hostname=quartz
datadir=/usr/local/mysql/data

[NDBD]
id=3
hostname=amber
datadir=/usr/local/mysql/data

[MYSQLD]
id=4
hostname=jade

[MYSQLD]
id=5
hostname=coral

[MYSQLD]
id=6
hostname=quartz

[MYSQLD]
id=7
hostname=amber

[NDB_MGMD]
id=8
hostname=ruby
datadir=/var/lib/mysql-cluster

Here is the my.cnf file for jade, coral, amber and quartz:

[MYSQLD]
ndbcluster
ndb-connectstring="emerald,ruby"

[MYSQL_CLUSTER]
ndb-connectstring="emerald,ruby"


Everything works as you would expect:

1) The data nodes and SQL nodes connect to whichever management server is available.
2) If both are started, both show the connections and they show a connection to each other.
3) Either can control any of the nodes and the changes are reflected in both right away in both.
4) Status messages are sent to both.

Note that the configuration files on both management nodes need to be the same.

-Joe



Edited 1 time(s). Last edit at 02/18/2006 10:38PM by Joseph Kouyoumjian.