We have a cluster of 3 Galera-MariaDB boxes, using InnoDB as an engine. Cluster's in sync, and reporting all good as far as I can see.
Much to my surprise, the size of the InnoDB data files differ quite a lot between the 3 boxes; some tables even are twice as big on some boxes.
That remains the case even after nuking the data from the box with the largest data files and letting it do a full SST to rejoin the cluster.
Any ideas what could explain this?