Symptom:
The Cache Engine 4 was reloaded in high traffic

In pang.log, before the restart, there is high io error. So after reach the threshold. CE restart as application detect abnormal i/o. 

28-03-14 12:17:26.999,  4124	[io:Warning]	int write_request_thread_c::write_to_block_device(fm_request_c*, bool&)	/dev/sdaq3 957:pwrite(107,0x21713e0d7,297, 490639998679) = -1 - Input/output error

28-03-14 13:55:28.623,  4103	[app:Warning]	int cache_out_c::handle_io_response(io_request_c*)	Connection #726944065 file 803DD6BFBD7D01356C3DB472B1C449D71DF03C57 volume 36 got READ_ERROR(6) response #91291617 - reason IO_R_WRONG_BUFF_OR_LEN state 1 asked for piece 0 : 8706904 -> 8912896 

In mg log,

Mar 28 14:33:48 ce-4 pang[24662]: replication layer is stuck or no spread is running , pang will restart 
Mar 28 14:33:48 ce-4 pang[24662]: kill spread!!!
Mar 28 14:39:28 ce-2 pang[27047]: Verifying blade 4 found mounted volumes or xfs_repair - will reboot it 
Mar 28 14:39:28 ce-2 pang[27047]: Going to execute reboot to blade 4...

Conditions:
The Cache Engine 4 was reloaded in high traffic