Postgresql creates high load on RHEV-M

Issue

postmaster services are in "D" state and the performance on server is very low, also there is high load average on server (8 cpu):

top - 17:13:06 up 63 days,  3:41,  1 user,  load average: 80.09, 46.42, 21.79
Tasks: 528 total,   3 running, 524 sleeping,   0 stopped,   1 zombie
Cpu(s): 67.9%us,  5.0%sy,  0.0%ni,  0.7%id, 26.0%wa,  0.0%hi,  0.4%si,  0.0%st
Mem:  32862812k total, 32651640k used,   211172k free,   168036k buffers
Swap: 16498684k total,   270532k used, 16228152k free, 12434972k cached

  PID USER      PR  NI  VIRT  RES  SHR S %CPU %MEM    TIME+  COMMAND                                                                                                                                                                          
19460 ovirt     20   0 13.1g 8.5g  22m S 306.2 27.2 999:08.69 java                                                                                                                                                                            
30735 postgres  20   0  278m 103m  37m S 27.4  0.3   2:16.97 postmaster                                                                                                                                                                       
22202 postgres  20   0  299m 124m  37m S 16.9  0.4  24:57.37 postmaster                                                                                                                                                                       
22219 postgres  20   0  298m 123m  37m S 14.9  0.4  23:36.22 postmaster
...

During the high load time / ("slash") file system utilization is changing frequently and the size of /var/lib/pgsql/data/base/pgsql_tmp is increasing.

Restarting postgresql service resolves the issue temporarily.