Comment 19 for bug 1243156

Funny, we have no "WSREP: BF lock wait long" in the log and locks a seldom. But we already had a cluster stall twice with quite a similar sympthoms: all nodes have many "wsrep in pre-commit stage" threads, then the whole cluster stalls. If you restart the "bad node" cluster unlocks. Only two differences: it happens not so often and there is no way to identify the "bad" node. Looks like "BF lock wait long" logging was disabled|removed while the problem is still there.