High disk utilization in RAID0 logic drive

[Thomas1313]

Hi,

I have HP ProLiant DL360 G6 server with integrated Smart Array P410i controller (no 512 MiB memory bank). Firmware version 6.40.

There are two logic drives:
Drive A (RAID0) - 2TB WD 2.5" SATA disk
Drive B (RAID0) - 2TB WD + 2TB Hitachi 2.5" SATA disks

OS Linux Debian 7.7 x64. Filesystem is ext4 for both logic drives (partitions created for maximum space).

This server is used as a file server. File operations like create/delete are very often. Write stream is about 10-15 MB/s. Used space is about 2.6 TB in several thousands files.

For the first time disk utilization was ~10% for both drives (measured by iostat) and the server worked fine. But in two weeks it became 100%!!! RAM grows rapidly and the system hangs up with no messages! e4defrag utility tells fragmentation only in a few files and there is no need to defragmentation.

Device: rrqm/s wrqm/s r/s w/s rkB/s wkB/s avgrq-sz avgqu-sz await r_await w_await svctm %util
sda 0.00 1.40 0.00 0.80 0.00 4.00 10.00 17.17 9018.00 0.00 9018.00 1250.00 100.00
sdb 0.00 30.20 0.00 44.60 0.00 4211.20 188.84 88.88 2872.59 0.00 2872.59 22.42 100.00

How can I reduce disk utilization? Is there any fragmentation and is it possible to do online defragmentation (I guessed ext4 filesystem do defragmentation automatically)? Is this a RAID controller issue? Please help!!!

P.S. This thread has been moved from ProLiant Servers (ML,DL,SL) to Linux > sysadmin. - Hp Forum Moderator