© 2004
Canonical Ltd.
•
Terms of use
•
Data privacy
•
Contact Launchpad Support
•
Blog
•
Careers
•
System status
•
a60fb26
(Get the code!)
Justice,
it shouldn't be the case that the whole cluster went down. The node that is inconsistent shuts down. (Note also that it doesn't crash, it shuts down by design.) Often it can happen that all nodes are inconsistent, so that all "slave" nodes shut down, but the "master" node, ie the one that you are executing the transaction against, is of course consistent with itself, so at least that one will keep on running.
Galera is designed to be tightly coupled like this on purpose, we really work hard to avoid data inconsistency. Other systems, notably MySQL replication, will allow nodes to be inconsistent, and of course that seemingly gives better availability but also it is then quite a common problem that your nodes can keep on diverging.