I'm having some trouble finding the cause of my slow raid1 array and I'm running out of tests I can run.
if I read from the drives individually(simultaneous) I get speeds around 55mb/s
if I read from the raid array I get 1.3 MB/s read and 3 MB/s write
fsck says my filesystem is clean
extensive smart-tests complete without error
memtest passed 6 times without error
(I have 1333 mhz ram installed, but for some reason If I force it on 1333 instead of auto it still shows up as 14XX in memtest)
I also get a "segfault (core dumped)" after a successfull sftp connection (the binary is fine, I checked the mdsum against sftp on my desktop) but I'm not sure this problem is related
mdadm - v3.3 - 3rd September 2013
Personalities : [raid1]
md0 : active raid1 sda1 sdb1
488255424 blocks super 1.2 [2/2] [UU]
bitmap: 0/4 pages [0KB], 65536KB chunk
unused devices: <none>
AMD E-450 PRO -> mobo & cpu
Sweex PU102 SATA Controller PCI -> this card is used for both drives
Model Family: Western Digital Caviar Green
Device Model: WDC WD5000AAVS-00ZTB0
Model Family: SAMSUNG SpinPoint T166
Device Model: SAMSUNG HD501LJ
both drives have 1 partition filling the entire disk
Any ideas would be greatly appreciated.
Using a bitmap is known to cause some slowdown but not as much as you report.
Try not using the controller card or check google for any reports of problems from other users.
Try to check the speed of each disk independently, then try both at the same time but not in raid. Check dmesg for any errors and check bios/addon card for any settings related to HDD, you probably want to use ahci mode and not legacy ide mode, but in any case try toggling that setting and/or any other(s) related to hard disk drives and see if they make any difference.
I performed the tests again after remounting the array and performance was normal.
(drives connected to the sweex card)
Here are the results:
UUID=1efaf26a-5035-42ef-b8bd-594583bc27cc /srv/backup ext4 rw,relatime,data=ordered,nosuid,nodev,noexec 0 2
/srv/backup /srv/nfs/backup none bind 0 0
dd if=/dev/zero of=largefile2 bs=1M count=5000
5000+0 records in
5000+0 records out
5242880000 bytes (5,2 GB) copied, 116,621 s, 45,0 MB/s
dd if=largefile2 of=/dev/null
10240000+0 records in
10240000+0 records out
5242880000 bytes (5,2 GB) copied, 87,425 s, 60,0 MB/s
hdparm -t /dev/sda
Timing buffered disk reads: 218 MB in 3.01 seconds = 72.36 MB/sec
hdparm -t /dev/sdb
Timing buffered disk reads: 228 MB in 3.02 seconds = 75.53 MB/sec
hdparm -t /dev/md0
Timing buffered disk reads: 210 MB in 3.01 seconds = 69.65 MB/sec
The results int the first post were obtained using dd.
I'll give the machine a reboot later on and see if the problem returns.
Anyway, thanks for the help.
Last edited by Midnightfox (2013-12-05 12:05:59)
After some more testing I have concluded the error must be sata cables or the sweex controller. I've tested the drives/array at different times and sometimes it's dead slow and other times it's the usual speed.