Threads_running becomes very large

btw 2^64-1 should be 2^32

I use Haproxy to loadbalance across 4 servers and sysbench to drive load:

sysbench --test=oltp --mysql-table-engine=innodb --oltp-table-size=1000000 --mysql-user=johan --mysql-password=123johan --mysql-host=127.0.0.1 --mysql-port=33306 --num-threads=64 --max-time=36000 --max-requests=0 run

on one node do:
service mysql restart
show global status like 'threads_running';

mysql> show global status like 'threads_running';
+-----------------+------------+
| Variable_name | Value |
+-----------------+------------+
| Threads_running | 4294967260 |
+-----------------+------------+
1 row in set (0.00 sec)

This value is the same across all servers.'

We are seeing this as well with Percona Xtradb-Cluster 5.5.24-23.6.340.rhel6 (2 nodes + 1 galera arbitrator) with an F5 bigIP load balancer. Jay Jansen noticed this was unusually high number and recommended we file a bug. Not sure it affects much but it definitely affects the ability to run pt-online-schema-change without setting absurdly high override values for --critical-load and --max-load. Notably we have not had a "node recovery" on our primary server as mentioned in the description above

mysql> show global status like 'threads_running';
+-----------------+------------+
| Variable_name | Value |
+-----------------+------------+
| Threads_running | 4294958896 |
+-----------------+------------+
1 row in set (0.00 sec)

Fix committed in lp:codership-mysql revno 3821 (http://bazaar.launchpad.net/~codership/codership-mysql/wsrep-5.5/revision/3821)