We run our application on WildFly 8.2 and it uses a non-XA MariaDb v10 datasource. We plan to bring high availability on all levels. So far we've achieved HA on the database side by running a Galera\MariaDB cluster. So far so good everything works well. Next we want to run WildFly in a cluster too. Galera cluster doesn't support XA transactions, so we have no choice but to stick with non-XA datasources on WildFly. No problem for a single WildFly.
But the big question is: is it possible to run WildFly in a cluster and only use non-XA datasources?
I'm not that familiar with Jboss/WildFly ( I'm the database guy in the equation). But based on this discussion Re: when to use xa-datasource I concluded that a XA datasources might be mandatory when running WildFly in a cluster. In such case we should rethink our HA approach on the database side.