I've been using MariaDB 10.2.6 with slave_run_triggers_for_rbr for some aggregation tables.
However, there's a big memory leak with this setup and the server regularly hangs.
Last time it hangs, innodb crash recovery could not start because MLOG_CHECKPOINT was not found.
My server has got a BBWC so any fsync should be persisted on disk. I dont understand how that could happen ?