Galera resource file issue

I'm using release v3.9.6.
I think release v3.9.6 have the same issue too.

The function

`detect_first_master`

Compare the last commit, to select best node.
If Node1 and Node2 as the same last_commit.
Then best_node have two different value.
Then Node1 and Node2 all the bootstrap node.
Then Node1 mysql got error:

`due to conflicting prims: older overrides`

And the last, Node1 status looks like fine, but can't INSERT or UPDATE, there raise
`1213, 'Deadlock found when trying to get lock; try restarting transaction'`

At that time, mysql wsrep status like this:
`+---------------+-------+`
`| Variable_name | Value |`
`+---------------+-------+`
`| wsrep_last_committed | 448630 |`
`| wsrep_incoming_addresses | 172.17.105.66:3306 |`
`| wsrep_cluster_status | Primary |`
`| wsrep_connected | ON |`
`| wsrep_ready | ON |`

There is node1 log:
`Feb 23 08:31:03 cvm66 galera(db)[5516]: INFO: Enter galera_start`
`Feb 23 08:31:04 cvm66 galera(db)[5516]: INFO: MySQL is not running`
`....`
`Feb 23 08:31:14 cvm66 galera(db)[5516]: INFO: attempting to detect last commit version`
`Feb 23 08:31:14 cvm66 galera(db)[5516]: INFO: Last commit version found: 448630`
`Feb 23 08:31:14 cvm66 galera(db)[5516]: INFO: Waiting on node <172.17.105.67> to report database status before Master instances can start.`
`Feb 23 08:31:15 cvm66 galera(db)[5516]: INFO: Leave galera_start 0`
`...`
`Feb 23 08:31:32 cvm66 pengine[4337]: notice: LogActions: Promote db:0#011(Slave -> Master 172.17.105.66)`
`Feb 23 08:31:32 cvm66 pengine[4337]: notice: LogActions: Promote db:1#011(Slave -> Master 172.17.105.67)`
`Feb 23 08:31:32 cvm66 galera(db)[6578]: INFO: Enter galera_promote`
`Feb 23 08:31:32 cvm66 galera(db)[6578]: INFO: Node <172.17.105.66> is bootstrapping the cluster`
`...`
`Feb 23 08:31:45 cvm66 galera(db)[6578]: INFO: Leave galera_promote 0`

There is node2 log:
`Feb 23 08:31:19 cvm67 galera(db)[5893]: INFO: Enter galera_start`
`Feb 23 08:31:20 cvm67 galera(db)[5893]: INFO: MySQL is not running`
`...`
`Feb 23 08:31:32 cvm67 galera(db)[5893]: INFO: attempting to detect last commit version`
`Feb 23 08:31:32 cvm67 galera(db)[5893]: INFO: Last commit version found: 448630`
`Feb 23 08:31:32 cvm67 galera(db)[5893]: INFO: Promoting 172.17.105.67 to be our bootstrap node`
`Feb 23 08:31:32 cvm67 galera(db)[5893]: INFO: Leave galera_start 0`
`...`
`Feb 23 08:31:32 cvm67 galera(db)[6452]: INFO: Enter galera_promote`
`Feb 23 08:31:32 cvm67 galera(db)[6452]: INFO: Node <172.17.105.67> is bootstrapping the cluster`
`...`
`Feb 23 08:31:46 cvm67 galera(db)[6452]: INFO: Bootstrap complete, promoting the rest of the galera instances.`
`Feb 23 08:31:46 cvm67 galera(db)[6452]: INFO: Leave galera_promote 0`

There is node1 mysql log:
`...`
`160223 8:31:55 [Warning] WSREP: a37e2635 conflicting prims: my prim: view_id(PRIM,a37e2635,11) other prim: view_id(PRIM,e76d9671,11)`
`...`
`160223 8:31:55 [ERROR] WSREP: exception from gcomm, backend must be restarted: a37e2635 aborting due to conflicting prims: older overrides (FATAL)`
`...`

Start by tracing the detect_first_master function in the Galera resource agent and compare its last-commit selection with the node1 and node2 logs. Reproduce or test the equal-last-commit case, then verify that only one node bootstraps and that conflicting primary views no longer occur.

Written by the indexing model from the issue text.