Multiple data center + failure/recovery options
So I've been able to set up and test all three configurations - MySQL Cluster, MySQL Galera master-master and MySQL replication - thanks to the online Configurator tool.
I have a couple of questions:
- Will the Configurator support multiple data center configurations soon? Or are we limited to building our own multi-data center configurations manually?
- I tried to shut down some nodes in all three scenarios (for failure/recovery testing) and then started them back up - both manually. But after re-starting, ClusterControl doesn't recognize the restarted nodes. How do we set up failure/recovery in ClusterControl - does that have to be done manually in the free version?
Sorry one more question:
3. Even after shutting down the data or MySQL nodes, I'm still able to connect to "mysql" on the ClusterControl and write and execute successful MySQL queries. Does that mean that ClusterControl has its own version/copy of MySQL that is handling the database queries and writing locally? How does that work?
That is great to hear!
1) Actually users have deployed Galera over multiple data centers, so it is possible but we havent written so much about how to do this really.
2 and 3) Do you think it would be possible to setup a remote session and try to see what is going on, as recovery is built into the free version.
You can send me an email to
and we can schedule a time for this.
One problem with EC2 is that sometimes when instances are rebooted they get another IP. That may cause a lot of problems and we cannot handle the IP/hostname change.
Read more at http://support.rightscale.com/06-FAQs/FAQ_0160_-_Can_the_public_or_private_IP_addresses_change_on_an_EC2_instance%3F
That's is why EC2 have ElasticIP. You can also try and use the Public DNS address of the instance when you enter hostname/ips in the Configurator, just set "Skip name resolve" to No.
I was using an m1.large instance, but I was also testing on several AWS regions, and perhaps some of them behave differently from the others. I'll probably try this again sometime next week and test it by restarting the instance rather than a stop/start. Thanks a lot for your help and quick responses, really appreciate it!
Please sign in to leave a comment.