HOWTO: Linux Software Raid using mdadm

[alecm3]

I decided to switch from SuSE to Ubuntu, and I am configuring my first production MySQL server with 8.04.
I am using software RAID 1.
The machine is not yet deployed, it's sitting in the office.
Suddenly, i hear loud fan noise, even though there are no tasks running, and I check the load average: 2.8! Using "top" I notice that
md2_resync process is the culprit.

Then I look at /proc/mdstat and see:

Personalities : [linear] [multipath] [raid0] [raid1] [raid6] [raid5] [raid4] [raid10]
md2 : active raid1 sda3[0] sdb3[1]
135347520 blocks [2/2] [UU]
[============>........] check = 62.9% (85253248/135347520) finish=7.9min speed=125122K/sec


So apparently, mdadm daemon decided to check the array.
How can i disable or control the schedule of these "checks" without having to disable mdadm daemon alltogether in /etc/init.d?

It's not acceptable for a production database machine to suddenly start a maintanence process that saturates its I/O bandwidth.

[bsmith1051]

I've never seen my array do that. Are you sure there wasn't a hw prob with one of the drives?

[AmJaD]

Here is alot of messing about with Ubuntu when it came to setting up raid and for only 2 HDDz you wont really see the difference. i hope in the next ubuntu version they make it as simple as some of the other distros.
A name like "Raiding Rabbit" will be cool

[bsmith1051]

for only 2 HDDz you wont really see the difference. i hope in the next ubuntu version they make it as simple as some of the other distros.

Which distros have simple/automatic support for RAID?

Also, in the process of configuring/testing that my RAID1 would boot degraded (by simply unplugging one drive) I found that the array booted-up and loaded OS, apps, etc about 25% faster than 1 drive alone. Not sure if that indicates an actual speed advantage to RAID1 (though it's supposed to intelligently distribute read requests between both drives) or if the degraded array simply ran slower than a 'normal' 1-drive system.

[alecm3]

I've never seen my array do that. Are you sure there wasn't a hw prob with one of the drives?

I doubt that, for several reasons:

1) When it does a rebuild, it looks like

[>....................] recovery = 0.1%

in my case, it was

[>....................] check = 0.1%

2) if there's a hardware problem, it will not try to check or even rebuild, it would simply mark disk as failed [U_].

3) I saw this thread http://lists.clug.org.za/pipermail/c...er/024059.html :

> Yes, current software RAID has an option to check the arrays periodically.
> Apparently you have this enabled. I'm not sure that checking involves
> re-syncing though. Perhaps there was some sort of problem found? If so it
> should be in your logs.

In my logs:

Aug 2 20:24:39 db2 -- MARK --
Aug 2 20:44:39 db2 -- MARK --
Aug 2 21:04:39 db2 -- MARK --
Aug 2 21:24:39 db2 -- MARK --
Aug 2 21:44:39 db2 -- MARK --
Aug 2 22:04:39 db2 -- MARK --
Aug 2 22:24:39 db2 -- MARK --
Aug 2 22:44:39 db2 -- MARK --
Aug 2 23:04:39 db2 -- MARK --
Aug 2 23:24:39 db2 -- MARK --
Aug 2 23:44:39 db2 -- MARK --
Aug 3 00:04:39 db2 -- MARK --
Aug 3 00:24:39 db2 -- MARK --
Aug 3 00:44:39 db2 -- MARK --
Aug 3 01:04:39 db2 -- MARK --
Aug 3 01:06:01 db2 kernel: [187105.414819] md: data-check of RAID array md0
Aug 3 01:06:01 db2 kernel: [187105.414824] md: minimum _guaranteed_ speed: 1000 KB/sec/disk.
Aug 3 01:06:01 db2 kernel: [187105.414826] md: using maximum available idle IO bandwidth (but not more than 200000 K
B/sec) for data-check.

No I/O error prior to this "data-check", so it seems quite sudden.

I wonder if this check is an addition to the newest version of mdadm (I am using Ubuntu 8.04 server with mdadm - v2.6.3 - 20th August 2007
I cannot find more information on this however.

[alecm3]

I doubt that, for several reasons:

1) When it does a rebuild, it looks like

[>....................] recovery = 0.1%

in my case, it was

[>....................] check = 0.1%

2) if there's a hardware problem, it will not try to check or even rebuild, it would simply mark disk as failed [U_].

3) I saw this thread http://lists.clug.org.za/pipermail/c...er/024059.html :

> Yes, current software RAID has an option to check the arrays periodically.
> Apparently you have this enabled. I'm not sure that checking involves
> re-syncing though. Perhaps there was some sort of problem found? If so it
> should be in your logs.

In my logs:

Aug 2 20:24:39 db2 -- MARK --
Aug 2 20:44:39 db2 -- MARK --
Aug 2 21:04:39 db2 -- MARK --
Aug 2 21:24:39 db2 -- MARK --
Aug 2 21:44:39 db2 -- MARK --
Aug 2 22:04:39 db2 -- MARK --
Aug 2 22:24:39 db2 -- MARK --
Aug 2 22:44:39 db2 -- MARK --
Aug 2 23:04:39 db2 -- MARK --
Aug 2 23:24:39 db2 -- MARK --
Aug 2 23:44:39 db2 -- MARK --
Aug 3 00:04:39 db2 -- MARK --
Aug 3 00:24:39 db2 -- MARK --
Aug 3 00:44:39 db2 -- MARK --
Aug 3 01:04:39 db2 -- MARK --
Aug 3 01:06:01 db2 kernel: [187105.414819] md: data-check of RAID array md0
Aug 3 01:06:01 db2 kernel: [187105.414824] md: minimum _guaranteed_ speed: 1000 KB/sec/disk.
Aug 3 01:06:01 db2 kernel: [187105.414826] md: using maximum available idle IO bandwidth (but not more than 200000 K
B/sec) for data-check.

No I/O error prior to this "data-check", so it seems quite sudden.

I wonder if this check is an addition to the newest version of mdadm (I am using Ubuntu 8.04 server with mdadm - v2.6.3 - 20th August 2007
I cannot find more information on this however.

This problem is listed as Ubuntu bug:

https://bugs.launchpad.net/ubuntu/+s...ux/+bug/212684

There is a cron script
/etc/cron.d/mdadm
which runs /usr/share/mdadm/checkarray on the first Sunday of each month (which happened yesterday).

root@db2:/etc/cron.d# more mdadm
#
# cron.d/mdadm -- schedules periodic redundancy checks of MD devices
#
# Copyright © martin f. krafft <madduck@madduck.net>
# distributed under the terms of the Artistic Licence 2.0
#
# $Id$
#

# By default, run at 01:06 on every Sunday, but do nothing unless the day of
# the month is less than or equal to 7. Thus, only run on the first Sunday of
# each month. crontab(5) sucks, unfortunately, in this regard; therefore this
# hack (see #380425).
6 1 * * 0 root [ -x /usr/share/mdadm/checkarray ] && [ $(date +\%d) -le 7 ] && /usr/share/mdadm/checkarray --cron --a
ll --quiet

Apparently, this cron killed the systems for quite a lot of people:
http://ubuntuforums.org/showthread.php?t=748418 etc

[alecm3]

Not sure if that indicates an actual speed advantage to RAID1 (though it's supposed to intelligently distribute read requests between both drives) or if the degraded array simply ran slower than a 'normal' 1-drive system.

Reads on RAID 1 with N disks are N times faster than on a single disk. Writes are the same speed or slightly slower.

[AmJaD]

Which distros have simple/automatic support for RAID?

Well there is fedora. that was painless when it came to setting up raids but for some reason i just didnt like it.

also with raid1... thats more mirroring (had drive back-up). i dont really see the point on linux based desktops. (i mean its not like your going to get attacked by a win32 worm is it).
Raid0 is more performace
I was running vista with x2 hdds but i really noticed a difference when i had 4 hdds together (ram bus speed playes a big part as well).

[crtlbreak]

Which distros have simple/automatic support for RAID?

I think simple & automatic are not necessarily RAID and nix approaches to anything - dont get me wrong I am not being critical of your comments or of the linux world - just a general approach of "rigid, manual and locked down" in the nix world versus automatic everything in the ms world.
just my observations.
regards

[crtlbreak]

.... Ubuntu, and I am configuring my first production MySQL server with 8.04. ....

Dear alecm3
Are you using Ubuntu Server? I know the respective kernels of desktop and server are different with server being more tweaked to perform server functionality as opposed to desktop.
Regards