Hi, it seems the problem of WSREP: BF lock wait long for trx .... is appear again and locking a whole cluster (galera). I have a cluster with 3 nodes using HAProxy and KeepAlived. The cluster hung several times for a week. And one of the node produced a bigger error log than the others.
As you can see inside attachment, the error log that i extracted before writing so many BF lock wait long for trx ....
If any mistake, pls let me know. I using earlier version of MariaDB, 10.1.31
is this still buggy? or i don't figure it out that this isn't bug?
Sorry for my bad english
*edit : For each hung, i take a look on the error log and seems that the crash triggered by sql statement "commit". Any idea?
OS = CentOS 7.2.1511
MariaDB = 10.1.31