Comment 7 for bug 1619740

the fail:

node-1.test.domain.local/mysqld.log:2016-09-02T02:45:04.537997+00:00 err: 2016-09-02 02:45:04 13875 [Warning] WSREP: TO isolation failed for: 3, schema: nova_api, sql: CREATE INDEX flavor_extra_specs_flavor_id_key_idx ON flavor_extra_specs (flavor_id, `key`). Check wsrep connection state and retry the query.

is happens right during the split-brain process:

node-1.test.domain.local/ocf-mysql-wss.log:496:2016-09-02T02:45:04.149886+00:00 err: ERROR: p_mysqld: check_if_galera_pc(): But I'm running a new cluster, PID:13875, this is a split-brain!

20 seconds later the galera cluster recovered:

node-1.test.domain.local/mysqld.log:2016-09-02T02:45:24.605658+00:00 err: 2016-09-02 02:45:24 17094 [Note] WSREP: New cluster view: global state: bf1c6c99-70b4-11e6-ad87-6a2b8c2d2a6a:965, view# 5: Primary, number of nodes: 3, my index: 2, protocol version 3
node-1.test.domain.local/mysqld.log:2016-09-02T02:45:24.606194+00:00 err: 2016-09-02 02:45:24 17094 [Note] WSREP: SST complete, seqno: 965

the cluster installed with fuel-ha-utils package version 9.0.0-1~u14.04+mos8568 which contain the fix for split-brain: https://review.openstack.org/#/c/361943/

all fine with mysql/galera but nova-manage don't handle deadlocks exeptions even the parameter nova/db_max_retries set to 20. nova-manage should retry in case of deadlock error before error is raised.