General Confusion (cluster crashing/unstable)

[Trey Stout]

Hello all. I have been wrestling with the new cluster for about 3 weeks now. I thought I had it figured out, but frequent crashing, and general instability are making me second guess my configuration.

Here is the setup:

4 machines total. 3 are webservers, 1 is a dedicated load balancer.
All are running 4.1.7 compiled with NDB from source on gentoo linux.
The load balancer has 2GB ram and the other 3 all have 4GB each.

The 3 webservers are running apache, and mysql. They all have their own on-disk data on each machine. But certain data needs to be global across all 4 of them.

I currently have 4 mgmd nodes, and 3 ndbd nodes. Here is the config.ini...

[NDBD DEFAULT]
NoOfReplicas=3
MaxNoOfConcurrentTransactions=8192
MaxNoOfConcurrentOperations=250000
IndexMemory=128M
DataMemory=384M

[MYSQLD DEFAULT]
[NDB_MGMD DEFAULT]
[TCP DEFAULT]

[NDB_MGMD]
HostName=192.168.4.5
[NDB_MGMD]
HostName=192.168.4.10
[NDB_MGMD]
HostName=192.168.4.11
[NDB_MGMD]
HostName=192.168.4.12

[NDBD]
HostName=192.168.4.10
DataDir= /huge/mysql-cluster
[NDBD]
HostName=192.168.4.11
DataDir= /huge/mysql-cluster
[NDBD]
HostName=192.168.4.12
DataDir= /huge/mysql-cluster

[MYSQLD]
[MYSQLD]
[MYSQLD]
[MYSQLD]
[MYSQLD]

I put that file on all 4 servers, and then start their ndb_mgmd instances. They all come up. Then I go to each server and start the ndbd instances. They all take a while, but eventually all show up in 'ndb_mgm show'.

mysqld on each server is set to connect to it's local copy of ndb_mgmd.

Please tell me I have something wrong? Is there a way to run a mysql API on each server without a ndb_mgmd process on each server? Am I configuring the nodes wrong? Is 4.1.7 just unstable? Please give me some clues!

Thanks a lot for reading all of this, I really want clustering to work, it could be a real life saver.