Hi there!
It seems like I have to ask for help, since this is really bothering me. After running successfully for a few months I have now experienced twice loosing all data of my clustered mysql databases. The scenario seems really odd - it seems to happen when the management node's config.ini is changed and might probably be me doing something the wrong way. According to this tutorial (which I used for the ultramonkey setup)
http://www.howtoforge.com/loadbalanced_mysql_cluster_debian_p2
you are supposed to run ndbd --initial when config.ini has changed on the management node. This might be entirely wrong. I have performed the following steps:
1) Stopped mysqld on each mysql node using mysqladmin
2) Stopped ndbd on each mysql node using 'killall ndbd'
3) Typed 'shutdown' on the management down in ndb_mgm
4) Edited config.ini (set the DataMemory and IndexMemory properties).
5) Started ndb_mgmd on the management node
6) Started ndbd --initial on each mysql node (according to the tutorial)
7) Started mysqld on each mysql node.
show; on the ndb_mgm shows everything is alright and here is where the fun begins. When starting the mysql client on any of the mysql nodes the databases can be found. Unfortunately they are COMPLETELY EMPTY - all data is lost, table definitions as well.
So what can be the problem ?! - the --initial in step 6 or ?! I'm unfortunately not able to play much around with this, because it's a production environment.
This is mysql 5.0.22 on linux 2.6.
Any help would be greatly appreciated - thanks for an excellent product !
Best regards,
Martin