If sometimes you got this error message from console:
14063.876009] BUG: soft lockup - CPU#0 stuck for 61s! klogd:896
http://14063.876022 Modules linked in: ipv6 binfmt_misc ext4 jbd2 crc16 dm_mirror rtc_core rtc_lib usbcore nls_base dm_region_hash dm_log ext3 jbd mbcache linear dm_mod xen_blkfront xen_netfront
http://14063.876051 Modules linked in: ipv6 binfmt_misc ext4 jbd2 crc16 dm_mirror rtc_core rtc_lib usbcore nls_base dm_region_hash dm_log ext3 jbd mbcache linear dm_mod xen_blkfront xen_netfront
This has nothing to do with the CPU issue or hardware issue.
Due to the way that ext4 delays allocations, you might be overwhelming EBS with a flood of writes. You might want to adjust how long the kernel accumulates dirty data destined to disk, or the amount of data it allows to be dirty. Note that this will change the performance characteristics of the system, so test this before applying to a production environment.
Try setting
vm.dirty_background_bytes = 100000000 vm.dirty_bytes = 100000000
in /etc/sysctl.conf and running "sysctl -p". Alternatively, try:
vm.dirty_expire_centisecs = 3000</code>