PXC crash

'perf top' output from high sys CPU nodes (I believe these are in a pre-crash state):

Events: 266K cycles
 27.18% [kernel] [k] _spin_lock
  6.44% mysqld [.] 0x54b935
  2.23% mysqld [.] ha_partition::change_partitions(st_ha_create_information*, char const*, unsigned l
  1.92% libc-2.12.so [.] memcpy
  1.63% mysqld [.] yaSSL::buildFinished(yaSSL::SSL&, yaSSL::Finished&, unsigned char const*)
  1.34% libc-2.12.so [.] _int_malloc

Another one (much higher sys):

Events: 632K cycles
 58.11% [kernel] [k] _spin_lock
  4.58% mysqld [.] my_regcomp
  3.22% mysqld [.] 0x54b935
  1.15% libc-2.12.so [.] _int_malloc
  0.93% libc-2.12.so [.] memcpy
  0.91% mysqld [.] yaSSL::buildFinished(yaSSL::SSL&, yaSSL::Finished&, unsigned char const*)
  0.78% mysqld [.] ha_partition::change_partitions(st_ha_create_information*, char const*, unsigned l
  0.76% mysqld [.] create_fromuni
  0.52% [kernel] [k] intel_idle
  0.50% mysqld [.] MYSQLparse(void*)
  0.50% mysqld [.] row_purge_remove_sec_if_poss_leaf
  0.49% [kernel] [k] schedule
  0.44% libc-2.12.so [.] _int_free
  0.41% mysqld [.] row_purge_node_create
  0.41% [kernel] [k] copy_user_generic_string
  0.38% mysqld [.] ha_partition::truncate_partition(Alter_info*, bool*)
  0.35% libc-2.12.so [.] malloc

top - 11:50:03 up 21 days, 20:12, 6 users, load average: 1.20, 1.10, 1.05
Tasks: 693 total, 1 running, 692 sleeping, 0 stopped, 0 zombie
Cpu(s): 24.4%us, 46.0%sy, 0.0%ni, 29.1%id, 0.0%wa, 0.0%hi, 0.5%si, 0.0%st
Mem: 132112432k total, 124533664k used, 7578768k free, 186180k buffers
Swap: 16777208k total, 4028k used, 16773180k free, 11420344k cached

  PID USER PR NI VIRT RES SHR S %CPU %MEM TIME+ COMMAND
16004 mysql 20 0 123g 107g 2.0g S 2259.2 85.4 54854:27 mysqld
 3265 root 20 0 124m 5100 1860 S 1.0 0.0 0:01.77 perl
 5577 root 20 0 15536 1736 944 R 1.0 0.0 0:01.01 top

Note: tried disabling transparent hugepages, no apparent effect:

cat /sys/kernel/mm/redhat_transparent_hugepage/enabled
always [never]

[root@db02 mysql]# cat /sys/block/sd*/queue/scheduler
[noop] anticipatory deadline cfq
[noop] anticipatory deadline cfq

sdb (second one) is the datadir. It is SSD drives.

Similar bug reported lp:1123233

Since there are 2 other bugs reported similar to this, following details are required:

a) If the tables have FK or not. -- This is present in description.

b) A core dump and/or backtrace

c) If not b), atleast the crash log from the error log.

b and c are required to check the nature of query (is it INSERT ... DUPLICATE?).

The other two bugs being lp:1149735 and lp:1123233