Raid0 array stresses only 1 disk out of 3


 
Thread Tools Search this Thread
Top Forums UNIX for Dummies Questions & Answers Raid0 array stresses only 1 disk out of 3
# 1  
Old 04-11-2016
Raid0 array stresses only 1 disk out of 3

Hi there,

I've setup a raid0 array of 3 identical disks using :
Code:
mdadm --create --verbose /dev/md0 --level=stripe --raid-devices=3 /dev/sdb1 /dev/sdc1 /dev/sdd1

I'm using dstat to monitor the disk activity :
Code:
dstat --epoch -D sdb,sdc,sdd --disk-util 30

The results show that the stress is not evenly split (stripped) across the 3 disks:
Code:
2016-04-11 09:35:30 |   26%   28%   27%
[...]
2016-04-11 10:15:00 |    0%  100%    0%
2016-04-11 10:15:30 |    0%    3%   97%
2016-04-11 10:16:00 |    0%    0%   81%
2016-04-11 10:16:30 |    0%    0%  100%
2016-04-11 10:17:00 |    0%    0%   30%
[...]
2016-04-11 11:28:30 |    0%    0%   55%
2016-04-11 11:29:00 |    0%    0%   49%
2016-04-11 11:29:30 |    0%    0%   31%
2016-04-11 11:30:00 |    0%    0%   73%
2016-04-11 11:30:30 |    0%    0%    4%
2016-04-11 11:31:00 |    0%    0%   99%
[...]
2016-04-11 11:32:00 |    0%    0%   81%
2016-04-11 11:32:30 |    0%    0%   43%
[...]
2016-04-11 11:43:30 |    0%   93%    0%
2016-04-11 11:44:00 |    0%  100%    0%
2016-04-11 11:44:30 |    0%   97%    0%
2016-04-11 11:45:00 |    0%  100%    0%
2016-04-11 11:45:30 |    0%   10%    0%
[...]
2016-04-11 11:51:30 |    0%   79%    0%
2016-04-11 11:52:00 |    0%  100%    0%
2016-04-11 11:52:30 |    1%    9%    1%
2016-04-11 11:53:00 |    0%  100%    0%
2016-04-11 11:53:30 |    0%   98%    0%
2016-04-11 11:54:00 |    0%   30%    0%
2016-04-11 11:54:30 |    1%    1%    1%
2016-04-11 11:55:00 |    2%    3%    2%
[...]
2016-04-11 12:07:30 |    0%   68%    1%
2016-04-11 12:08:00 |    0%  100%    0%
2016-04-11 12:08:30 |    0%  100%    0%
2016-04-11 12:09:00 |    0%   38%    0%
[...]
2016-04-11 12:23:00 |    0%   84%    1%
2016-04-11 12:23:30 |    0%   58%    0%
[...]
2016-04-11 14:17:00 |    0%   43%    0%
2016-04-11 14:17:30 |    0%   99%    0%
2016-04-11 14:18:00 |    0%  100%    0%
2016-04-11 14:18:30 |    1%    6%    1%
[...]
2016-04-11 14:46:30 |    2%    2%    1%
[...]
2016-04-11 14:48:00 |    1%    9%    1%
2016-04-11 14:48:30 |    0%  100%    0%
2016-04-11 14:49:00 |    0%   96%    0%
2016-04-11 14:49:30 |    0%  100%    0%
2016-04-11 14:50:00 |    0%   99%    0%
2016-04-11 14:50:30 |    0%  100%    0%
2016-04-11 14:51:00 |    0%   41%    0%
2016-04-11 14:51:30 |    0%  100%    0%
2016-04-11 14:52:00 |    2%   18%    2%
[...]
2016-04-11 15:23:30 |    3%    5%    3%
[...]
2016-04-12 09:25:30 |    4%    3%    3%

Do you have an explanation?
Thanks for your help.

Santiago

OS : Debian Wheezy 7.4
Disks : ATA Hitachi HUA72302, 2000GB

Last edited by chebarbudo; 04-12-2016 at 08:00 AM.. Reason: edited command
# 2  
Old 04-11-2016
Hi,

This could be a number of things, but it will most likely revolve around the stripe size.

Regards

Gull04
# 3  
Old 04-12-2016
Hi Gull04,

Thank you for your answer.
Is "stripe size" the same as "chunk size"?

Apparently, mine is 512k:
Code:
cat /proc/mdstat

returns
Code:
Personalities : [raid0]
md0 : active raid0 sdd1[2] sdc1[1] sdb1[0]
      5860543488 blocks super 1.2 512k chunks

unused devices: <none>

How can I identify if this is the source of the problem?

Regards
Santiago
# 4  
Old 04-12-2016
RAID0 "stripes" the data across the three actuators you have and the stripe size (that's official RAID speak) is the minimum allocation. So if the stripe is 2k then the first 2k bytes of a file is written to the first drive, the next 2k to the second drive, and the third 2k to the third drive. It then goes back to the first drive, and so on.

So it's not difficult to see that writing lots of small files will give unpredictable results respecially if they're less than 2k each. Also, read requests can only be satisfied be reading the drive(s) where the files were written.

So your results are misleading.

If you have a desire to test this then you need to do something like......
Create a 4GB file on (ideally) an internal drive not part of this RAID0 array. Kick all the users off if you can and then copy this 4GB to the RAID filesystem and take your measurements whilst that's going on. It won't be precise but should give you a better set of figures.
# 5  
Old 04-12-2016
Quote:
Originally Posted by hicksd8
Create a 4GB file on (ideally) an internal drive not part of this RAID0 array.
Wouldn't it be sufficient to fire 4GB worth of any data (for instance some brand new hexadecimal zeroes freshly out of /dev/zero) with dd? Like

Code:
dd if=/dev/zero of=/the/raid/somefile bs=1G count=4

True, this will be off by the overhead of /dev/zero, wouldn't that be negligible given the bandwidth of disks and the memory interface (which are apart some orders of magnitude)?

I hope this helps.

bakunin
This User Gave Thanks to bakunin For This Post:
# 6  
Old 04-12-2016
@bakunin......point taken.....good idea.

Sent from my HTC Desire S using Tapatalk 2
# 7  
Old 04-14-2016
Hi guys,

Thank you very much for your contributions.

First of all, my problem does not happen any more. I created the raid with sdb, sdc and sdd on April 11 at 09:35.
Until 11:32, sdd was very busy, then until 14:51, sdc was very busy.
Since then (3 days), the 3 disks are always under the same moderate load altogether (0-20%). The server is used by 5 graphic designers manipulating quite large files (100M-2G).

I ran some tests and the results leave me quite puzzled. So I created simultaneously 10 files. 1GB each. But all the load went on sda. Leaving sdb, sdc and sdd with a moderate 20% load.

The command:
Code:
for i in {1..10}; do
  file=$(mktemp /galaxy/XXXXXXX)
  echo $file >> /galaxy/dd.files
  dd if=/dev/zero of=$file bs=1G count=1 &
  echo $!    >> /galaxy/dd.pids
done

The output of dstat:
Code:
----system---- sda--sdb--sdc--sdd-
     time     |util:util:util:util
14-04 15:56:30|  21:   0:   0:   0
14-04 15:57:00| 100:   0:   0:   0
14-04 15:57:30| 101:   0:   0:   0
14-04 15:58:00| 100:   2:   2:   1
14-04 15:58:30| 101:   3:   4:   2
14-04 15:59:00| 102:   4:   5:   4
14-04 15:59:30|  98:   2:   3:   2
14-04 16:00:00| 100:   4:   4:   2
14-04 16:00:30| 103:  16:  16:  15
14-04 16:01:00|  98:  16:  17:  15
14-04 16:01:30| 101:  15:  15:  15
14-04 16:02:00|  99:   9:   8:   8
14-04 16:02:30| 100:   3:   4:   3
14-04 16:03:00| 100:   2:   4:   3
14-04 16:03:30| 104:   4:   4:   3
14-04 16:04:00|  95:   4:   4:   3
14-04 16:04:30| 100:   3:   4:   2
14-04 16:05:00| 101:   3:   4:   3
14-04 16:05:30|  99:  12:  13:  12
14-04 16:06:00| 102:  20:  22:  18
14-04 16:06:30|  98:  17:  19:  18
14-04 16:07:00| 101:   7:   9:   8
14-04 16:07:30|  99:   4:   5:   3
14-04 16:08:00| 102:   4:   5:   3
14-04 16:08:30|  98:   3:   5:   3
14-04 16:09:00| 100:   5:   7:   5
14-04 16:09:30| 101:   5:   5:   4
14-04 16:10:00| 100:   4:   4:   2
14-04 16:10:30| 100:  17:  18:  16
14-04 16:11:01| 105:  16:  20:  16
14-04 16:11:30|  95:  15:  17:  17
14-04 16:12:00| 100:  12:  11:  10
14-04 16:12:30|  34:  15:  16:  14

Is /dev/zero an actual file of sda?
How do you interpret the results?

Regards
Santiago
 
Login or Register to Ask a Question

Previous Thread | Next Thread

9 More Discussions You Might Find Interesting

1. AIX

DISK ARRAY PROTECTION SUSPENDED message following disk replacement

Hello, On 4/20/2018, we performed a disk replacement on our IBM 8202 P7 server. After the disk was rebuilt, the SAS Disk Array sissas0 showed a status of degraded. However, the pdisks in the array all show a status of active. We did see a message in errpt. DISK ARRAY PROTECTION SUSPENDED. ... (3 Replies)
Discussion started by: terrya
3 Replies

2. Filesystems, Disks and Memory

DISK ARRAY PROTECTION SUSPENDED message displayed following disk replacement

Hello, On 4/20/2018, we performed a disk replacement on our IBM 8202 P7 server. After the disk was rebuilt, the SAS Disk Array sissas0 showed a status of degraded. However, the pdisks in the array all show a status of active. We did see a message in errpt. DISK ARRAY PROTECTION SUSPENDED. ... (1 Reply)
Discussion started by: terrya
1 Replies

3. UNIX for Advanced & Expert Users

Disk Array

HI: I have a DUAL redundant system running in HP-UX 10.20 ( 2 servers) Both servers share an external SCSI disk array. Both server runs a Data base wich uses the disk array to write the data, and that is the way they share the information. The way that the servers see the disk array, is like... (1 Reply)
Discussion started by: pmoren
1 Replies

4. Solaris

Disk Array

So I have 2 solaris sun servers one with a disk array with 14 drives which I need to move to the other server. How do I move the disk array configured as a Raid 5 to another server without losing data? So far I know I'll connect the drive, run devfsadm, use format to verify the server see's... (5 Replies)
Discussion started by: Kjons76
5 Replies

5. UNIX for Dummies Questions & Answers

Why is RAID0 faster?

I have read anecdotes about people installing RAID0 (RAID - Wikipedia, the free encyclopedia) on some of their machines because it gives a performance boost. Because bandwidth on the motherboard is limited, can someone explain exactly why it should be faster? (7 Replies)
Discussion started by: figaro
7 Replies

6. Linux

Raid0 recovery from external HD

Hi, I have the Lacie Big Disk, which is a external hard drive enclosure in a hardware RAID0 array of 2x250GB disks. The RAID controller seems to have died, leaving me with 2 working hard drives but no way to get the data. I tried hooking the drives up to a windows machine and using Raid... (4 Replies)
Discussion started by: dangral
4 Replies

7. UNIX for Advanced & Expert Users

3510 Disk Array Problem

I have a 3510 disk array attached to a T2000 server. The dmesg command shows disk error as follows and is generated a couple of times during the day Aug 18 03:35:51 myserver SUNWscsdMonitor: <rctrl6042> Standard General Event, CHL:2 ID:22 Drive NOTICE: Drive Recovered Error - 5F8E1F... (1 Reply)
Discussion started by: Tirmazi
1 Replies

8. Solaris

Solaris RAID0 doubt...

friends, Suppose I am typing metastat command and it is showing: d100: Concat/Stripe Size: 369495 blocks (180 MB) Stripe 0: (interlace: 32 blocks) Device Start Block Dbase Reloc c1d0s0 16065 Yes Yes c1d0s1 0 No Yes... (4 Replies)
Discussion started by: saagar
4 Replies

9. Solaris

A1000 Disk storage array

I am new to the unix world. I have SunBlade 100 and A1000 Disk storage array with 12 Hard drives. I used SCSI card and SCSI cables to connect. When I do the format command,I can see disk storage as one disk instead of 12 disks as below. Could anybody can explain why? What should I do in order... (1 Reply)
Discussion started by: Dulasi
1 Replies
Login or Register to Ask a Question

Featured Tech Videos