MySQL Cluster: Changing datamemory requires node restart with –initial?

I ran into something with mysql cluster today which boggles my mind.  On page, it is documented that if we you change datamemory parameter in the config.ini under mysql-cluster like below, you would have to restart nodes to reread the configuration.

HostName=   # the IP of the first data node

But when I tried the: 2 restart in the mysql cluster managment console (ndb_mgm), it didn’t re-read the new configuration.  DataMemory was initially set to 2G and it continued using that configuration.  I even tried: 2 stop, and than logged on the node server and started ndbd manually, but it refused to read the new configuration.  It is only when I did ndbd –initial, nodes read in the configuration and were able to use more than 2 gigs of ram.

Did anybody else run into this?  Am I reading the documentation on MySQL site incorrectly?

4 Responses to “ MySQL Cluster: Changing datamemory requires node restart with –initial? ”

  1. November 13th, 2007 | 8:11 pm

    ndb_mgmd reads config on startup. So you need to restart the management server first. Then a restart of data nodes (not initial)

  2. November 13th, 2007 | 8:26 pm

    As an additional note, you need to restart all of the management servers – and they all need to be down at the same time at some point, because ndb_mgmd will get its config from any running management servers.

  3. November 14th, 2007 | 9:04 am

    I understand why this has to be, but this kind of thing is why we decided cluster was not ready for us. If you have a nice static predictable environment that will only need to change once a year, cluster is great. It scales awesomely. But we are moving to fast and decided to just do replication which does not scale up as well but will scale out much better and faster with no down time.

  4. c0ld
    November 22nd, 2007 | 9:23 pm

    Thank you very much. Very informative.

Leave a reply

To prove that you're not a bot, enter this code
Anti-Spam Image