we are running a MariaDB Galera Cluster on Debian Jessie and we are sometimes experiencing the issue analyzed here: https://serverfault.com/a/715940/115361
I suspect it happens when systemd-journald is being restarted while everything else is running.
The immediately visible hint to that is when systemctl status mariadb.service on a node does not show any recent logging output anymore (although there should be some).
mysqld is still running normally, accepting queries and replicating to the other nodes, but an SST that tries to use such a defective node as donor will always fail with Will never receive state. Need to abort.
Restarting mariadb.service on the defective donor node fixes the issue until the logger process dies the next time.
mariadb-service-convert created the dropin configuration file to make mysqld log directly to syslog:
I'd be happy to assist with more information to solve this issue soon.