Our MariaDB galera cluster (on RHEL) runs with
[mysqld]
max_allowed_packet = 16M
slow_query_log = 1
slow_query_log_file = /var/log/mariadb/slow-query.log
long_query_time = 10
[galera]
binlog_format=ROW
default-storage-engine=innodb
innodb_autoinc_lock_mode=2
query_cache_size=16M
query_cache_type=0
innodb_flush_log_at_trx_commit=0
innodb_buffer_pool_size=256M
#Galera settings
wsrep_on=ON
wsrep_provider=/usr/lib64/galera/libgalera_smm.so
wsrep_sst_method=rsync
wsrep_slave_threads=4
taken/adjusted from Galera MySQL cluster with VIPs and HAProxy for IDO-Mysql and more
This ran pretty well, until we moved to a postgres setup when we switched to IcingaDB. With wsrep_sst_method=mariabackup we had performance and reliability issues though. Switching to rsync fixed that pretty much. No cluster rebuild since then, afair.
Now only the Director and some module DBs remain, though.
Maybe it helps you in some way 