We use a 3 node PXC 5.7 Cluster as our database backend and I guess, we have a replication problem. When I look at PMMs “MySQL Table Statistics” and compare the “Largest Tables by Size”, there is one node which has a quite different order for this list.
Here are two screenshot, first ist our only write node, second ist one of the nodes where it is replicated to. How could theses tables have different sizes? And this only happens on this one node.
Run pt-table-checksum to produce a consistency check on this cluster.
On Using pt-table-checksum with Percona XtraDB Cluster you can see how to do it.
The article How to Handle pt-table-checksum Errors can be useful if you have problems to run pt-table-checksum.
Let us know how it goes.
@reventix_gmbh The stats you see in PMM come from tables in MySQL that are not always up-to-date. As eder_santos pointed out, use pt-table-cheksum to verify all the row data is the same.
You could also run OPTIMIZE TABLE to force the tables to rebuild and then the stats would match.