RAID5 on different sized disks on low-end machine

RAID5 on different sized disks on low-end machine

[Derek Piper]

Hi,

I am new to RAID / md devices, although I've used Linux for a number
of years. I decided it was high-time I had a RAID at home for
important things (email, web-sites, son's baby pics, mp3s etc.). I
happen to have a 3 Seagate 60GB hds and 1 80GB Seagate hd that I am
considering using for a RAID.

My question is this, is it possible (and even a good idea) to use all
4 hard drives as members of a 4 x 60GB RAID5 array by leaving 20GB of
the 80GB drive as a non-raided partition? I'll be using a Promise
Ultra TX2/100 controller.

i.e.

hde -> 60
hdf -> 60
hdg -> 60
hdh -> 60/20

I heard about RAID6 too, though I'm assuming that will use up another
disk's worth of disk space too.

i.e. RAID5 = 180GB usable size,wherease RAID6 = 120GB .. am I correct
in my thinking?

I know many of you use far larger hard drives, I'm just trying to use
the components I already had spare from a number of machines and
reorganize to a RAID-backed fileserver.

The machine is a dual pentium-pro 200 (320MB RAM) .. would that be a
dumb idea to use RAID5 on it because of the parity calculations
needed?

Further to that, would it be a smarter idea to use RAID1 on all 4 of
some small partition(s) at the start of the disks to house
boot/root/usr partitions, and only RAID5 on a larger 'data' area of
the drive that is more likely to be read than written to?

Comments are appreciated.

Thanks,

Derek

Re: RAID5 on different sized disks on low-end machine

[Maarten]

On Tuesday 11 January 2005 19:47, Derek Piper wrote:

> My question is this, is it possible (and even a good idea) to use all
> 4 hard drives as members of a 4 x 60GB RAID5 array by leaving 20GB of
> the 80GB drive as a non-raided partition? I'll be using a Promise
> Ultra TX2/100 controller.

That's perfectly okay.

> i.e. RAID5 = 180GB usable size,wherease RAID6 = 120GB .. am I correct
> in my thinking?

Yes.

> I know many of you use far larger hard drives, I'm just trying to use
> the components I already had spare from a number of machines and
> reorganize to a RAID-backed fileserver.

My first raid was a raid-0 from two 4.5 GB scsi disks...
My first raid at home was on 4x40GB drives, raid5.

> The machine is a dual pentium-pro 200 (320MB RAM) .. would that be a
> dumb idea to use RAID5 on it because of the parity calculations
> needed?

That raid-5 array above ran on a K6-300...  may not be fast but you can almost 
be sure that a 100Mbit ethernet is still slower...

> Further to that, would it be a smarter idea to use RAID1 on all 4 of
> some small partition(s) at the start of the disks to house
> boot/root/usr partitions, and only RAID5 on a larger 'data' area of
> the drive that is more likely to be read than written to?

YES !  If only because it is very hard to boot from raid-5.
But you could use that spare 20GB for the OS, couldn't you ?
Unless you want that redundant too.

Maarten

RAID5 on different sized disks on low-end machine

[Derek Piper]

Wow, you replied so quick to my first posting already, thanks Maarten! :)

On Tue, 11 Jan 2005 20:10:05 +0100, Maarten <maarten@ultratux.net> wrote:
> On Tuesday 11 January 2005 19:47, Derek Piper wrote:
>
> > My question is this, is it possible (and even a good idea) to use all
> > 4 hard drives as members of a 4 x 60GB RAID5 array by leaving 20GB of
> > the 80GB drive as a non-raided partition? I'll be using a Promise
> > Ultra TX2/100 controller.
>
> That's perfectly okay.

Thank you, good to know. My plan kinda hinges on that. :)

>
> > i.e. RAID5 = 180GB usable size,wherease RAID6 = 120GB .. am I correct
> > in my thinking?
>
> Yes.

Thanks, I've only just started looking at what RAID6 offers. For
bigger arrays than mine I think.

> > I know many of you use far larger hard drives, I'm just trying to use
> > the components I already had spare from a number of machines and
> > reorganize to a RAID-backed fileserver.
>
> My first raid was a raid-0 from two 4.5 GB scsi disks...
> My first raid at home was on 4x40GB drives, raid5.
>
> > The machine is a dual pentium-pro 200 (320MB RAM) .. would that be a
> > dumb idea to use RAID5 on it because of the parity calculations
> > needed?
>
> That raid-5 array above ran on a K6-300...  may not be fast but you can almost
> be sure that a 100Mbit ethernet is still slower...

Okay, that's good to hear.. thank you :)

> > Further to that, would it be a smarter idea to use RAID1 on all 4 of
> > some small partition(s) at the start of the disks to house
> > boot/root/usr partitions, and only RAID5 on a larger 'data' area of
> > the drive that is more likely to be read than written to?
>
> YES !  If only because it is very hard to boot from raid-5.
> But you could use that spare 20GB for the OS, couldn't you ?
> Unless you want that redundant too.

Ah, that's a good point about booting. Yea, I'd want to be able to get
back running as simply as possible if a disk failure happens.

The extra 20GB I was thinking I could use for something un-important,
like downloads, temporary 'scratch' space etc. Maybe squid cache.

Thanks for getting back with me so quick. Nice to bounce ideas off people.

Derek

--
Derek Piper - derek.piper@gmail.com
http://doofer.org/

Re: RAID5 on different sized disks on low-end machine

[Derek Piper]

Hi,

I revised my idea and thought about RAID 1+0 for some partitions,
since there are 4 drives. This outline below might clarify what I was
trying to mention earlier. Is this a feasible set-up that would be
bootable (kernel compiled-in md, I'm no stranger to compiling
kernels)? I'm interested to hear comments/opinions since I've never
done this before. Like I said, it'll be running on a Dual-pentium pro
200 (W6-LI) machine, I have no idea if machines of that vintage have
the 'cojones' for software raid or not.

My ideas of RAID1+0 / RAID5 disk system partitions							
		MB					
/dev/hde	60GB	57241	(from controller)				
/dev/hdf	60GB	57241	(from controller)				
/dev/hdg	60GB	57241	(from controller)				
/dev/hdh	80GB	78125	(unconfirmed)				
							
/dev/hd* = applies to all drives considered here							
							
Device	MB	Type	GB	Mountpoint	MD device	RAIDed size (MB)	GB
/dev/hd*1	20	RAID1 + 0	0.02	/boot	/dev/md1	40	0.04
/dev/hd*2	192	RAID1 + 0	0.19	Swap	/dev/md2	384	0.38
/dev/hd*5	2048	RAID1 + 0	2	/	/dev/md5	4096	4
/dev/hd*6	2048	RAID5	2	/home	/dev/md6	6144	6
/dev/hd*7	52933	RAID5	51.69	/data	/dev/md7	158799	155.08

Does swap being raided make sense? I hear that sometimes it's a good
idea since a disk failure won't make you crash and then I heard
elsewhere that it doesn't matter and the kernel automatically raids
swap partitions anyway. I prepared the above in a spreadsheet btw so I
could work out partition sizes.

Thanks in advance again for any comments.

Derek

On Tue, 11 Jan 2005 13:47:20 -0500, Derek Piper <derek.piper@gmail.com> wrote:
> Hi,
> 
> I am new to RAID / md devices, although I've used Linux for a number
> of years. I decided it was high-time I had a RAID at home for
> important things (email, web-sites, son's baby pics, mp3s etc.). I
> happen to have a 3 Seagate 60GB hds and 1 80GB Seagate hd that I am
> considering using for a RAID.
> 
> My question is this, is it possible (and even a good idea) to use all
> 4 hard drives as members of a 4 x 60GB RAID5 array by leaving 20GB of
> the 80GB drive as a non-raided partition? I'll be using a Promise
> Ultra TX2/100 controller.
> 
> i.e.
> 
> hde -> 60
> hdf -> 60
> hdg -> 60
> hdh -> 60/20
> 
> I heard about RAID6 too, though I'm assuming that will use up another
> disk's worth of disk space too.
> 
> i.e. RAID5 = 180GB usable size,wherease RAID6 = 120GB .. am I correct
> in my thinking?
> 
> I know many of you use far larger hard drives, I'm just trying to use
> the components I already had spare from a number of machines and
> reorganize to a RAID-backed fileserver.
> 
> The machine is a dual pentium-pro 200 (320MB RAM) .. would that be a
> dumb idea to use RAID5 on it because of the parity calculations
> needed?
> 
> Further to that, would it be a smarter idea to use RAID1 on all 4 of
> some small partition(s) at the start of the disks to house
> boot/root/usr partitions, and only RAID5 on a larger 'data' area of
> the drive that is more likely to be read than written to?
> 
> Comments are appreciated.
> 
> Thanks,
> 
> Derek
> 


-- 
Derek Piper - derek.piper@gmail.com
http://doofer.org/

RE: RAID5 on different sized disks on low-end machine

[Guy]

As far as I know, you can only boot from RAID1.

Guy

-----Original Message-----
From: linux-raid-owner@vger.kernel.org
[mailto:linux-raid-owner@vger.kernel.org] On Behalf Of Derek Piper
Sent: Tuesday, January 11, 2005 2:14 PM
To: linux-raid@vger.kernel.org
Subject: Re: RAID5 on different sized disks on low-end machine

Hi,

I revised my idea and thought about RAID 1+0 for some partitions,
since there are 4 drives. This outline below might clarify what I was
trying to mention earlier. Is this a feasible set-up that would be
bootable (kernel compiled-in md, I'm no stranger to compiling
kernels)? I'm interested to hear comments/opinions since I've never
done this before. Like I said, it'll be running on a Dual-pentium pro
200 (W6-LI) machine, I have no idea if machines of that vintage have
the 'cojones' for software raid or not.

My ideas of RAID1+0 / RAID5 disk system partitions

		MB					
/dev/hde	60GB	57241	(from controller)

/dev/hdf	60GB	57241	(from controller)

/dev/hdg	60GB	57241	(from controller)

/dev/hdh	80GB	78125	(unconfirmed)				
							
/dev/hd* = applies to all drives considered here

							
Device	MB	Type	GB	Mountpoint	MD device	RAIDed size
(MB)	GB
/dev/hd*1	20	RAID1 + 0	0.02	/boot	/dev/md1	40
0.04
/dev/hd*2	192	RAID1 + 0	0.19	Swap	/dev/md2	384
0.38
/dev/hd*5	2048	RAID1 + 0	2	/	/dev/md5	4096
4
/dev/hd*6	2048	RAID5	2	/home	/dev/md6	6144	6
/dev/hd*7	52933	RAID5	51.69	/data	/dev/md7	158799
155.08

Does swap being raided make sense? I hear that sometimes it's a good
idea since a disk failure won't make you crash and then I heard
elsewhere that it doesn't matter and the kernel automatically raids
swap partitions anyway. I prepared the above in a spreadsheet btw so I
could work out partition sizes.

Thanks in advance again for any comments.

Derek

On Tue, 11 Jan 2005 13:47:20 -0500, Derek Piper <derek.piper@gmail.com>
wrote:
> Hi,
> 
> I am new to RAID / md devices, although I've used Linux for a number
> of years. I decided it was high-time I had a RAID at home for
> important things (email, web-sites, son's baby pics, mp3s etc.). I
> happen to have a 3 Seagate 60GB hds and 1 80GB Seagate hd that I am
> considering using for a RAID.
> 
> My question is this, is it possible (and even a good idea) to use all
> 4 hard drives as members of a 4 x 60GB RAID5 array by leaving 20GB of
> the 80GB drive as a non-raided partition? I'll be using a Promise
> Ultra TX2/100 controller.
> 
> i.e.
> 
> hde -> 60
> hdf -> 60
> hdg -> 60
> hdh -> 60/20
> 
> I heard about RAID6 too, though I'm assuming that will use up another
> disk's worth of disk space too.
> 
> i.e. RAID5 = 180GB usable size,wherease RAID6 = 120GB .. am I correct
> in my thinking?
> 
> I know many of you use far larger hard drives, I'm just trying to use
> the components I already had spare from a number of machines and
> reorganize to a RAID-backed fileserver.
> 
> The machine is a dual pentium-pro 200 (320MB RAM) .. would that be a
> dumb idea to use RAID5 on it because of the parity calculations
> needed?
> 
> Further to that, would it be a smarter idea to use RAID1 on all 4 of
> some small partition(s) at the start of the disks to house
> boot/root/usr partitions, and only RAID5 on a larger 'data' area of
> the drive that is more likely to be read than written to?
> 
> Comments are appreciated.
> 
> Thanks,
> 
> Derek
> 


-- 
Derek Piper - derek.piper@gmail.com
http://doofer.org/
-
To unsubscribe from this list: send the line "unsubscribe linux-raid" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html

Re: RAID5 on different sized disks on low-end machine

[Robin Bowes]

Derek Piper wrote:
> Hi,
> 
> I revised my idea and thought about RAID 1+0 for some partitions,
> since there are 4 drives. This outline below might clarify what I was
> trying to mention earlier. Is this a feasible set-up that would be
> bootable (kernel compiled-in md, I'm no stranger to compiling
> kernels)? I'm interested to hear comments/opinions since I've never
> done this before. Like I said, it'll be running on a Dual-pentium pro
> 200 (W6-LI) machine, I have no idea if machines of that vintage have
> the 'cojones' for software raid or not.
> 
> My ideas of RAID1+0 / RAID5 disk system partitions							
> 		MB					
> /dev/hde	60GB	57241	(from controller)				
> /dev/hdf	60GB	57241	(from controller)				
> /dev/hdg	60GB	57241	(from controller)				
> /dev/hdh	80GB	78125	(unconfirmed)				
> 							
> /dev/hd* = applies to all drives considered here							
> 							
> Device	MB	Type	GB	Mountpoint	MD device	RAIDed size (MB)	GB
> /dev/hd*1	20	RAID1 + 0	0.02	/boot	/dev/md1	40	0.04
> /dev/hd*2	192	RAID1 + 0	0.19	Swap	/dev/md2	384	0.38
> /dev/hd*5	2048	RAID1 + 0	2	/	/dev/md5	4096	4
> /dev/hd*6	2048	RAID5	2	/home	/dev/md6	6144	6
> /dev/hd*7	52933	RAID5	51.69	/data	/dev/md7	158799	155.08
> 
> Does swap being raided make sense? I hear that sometimes it's a good
> idea since a disk failure won't make you crash and then I heard
> elsewhere that it doesn't matter and the kernel automatically raids
> swap partitions anyway. I prepared the above in a spreadsheet btw so I
> could work out partition sizes.
> 
> Thanks in advance again for any comments.
> 
> Derek
> 
> On Tue, 11 Jan 2005 13:47:20 -0500, Derek Piper <derek.piper@gmail.com> wrote:
> 
>>Hi,
>>
>>I am new to RAID / md devices, although I've used Linux for a number
>>of years. I decided it was high-time I had a RAID at home for
>>important things (email, web-sites, son's baby pics, mp3s etc.). I
>>happen to have a 3 Seagate 60GB hds and 1 80GB Seagate hd that I am
>>considering using for a RAID.
>>
>>My question is this, is it possible (and even a good idea) to use all
>>4 hard drives as members of a 4 x 60GB RAID5 array by leaving 20GB of
>>the 80GB drive as a non-raided partition? I'll be using a Promise
>>Ultra TX2/100 controller.
>>
>>i.e.
>>
>>hde -> 60
>>hdf -> 60
>>hdg -> 60
>>hdh -> 60/20
>>
>>I heard about RAID6 too, though I'm assuming that will use up another
>>disk's worth of disk space too.
>>
>>i.e. RAID5 = 180GB usable size,wherease RAID6 = 120GB .. am I correct
>>in my thinking?
>>
>>I know many of you use far larger hard drives, I'm just trying to use
>>the components I already had spare from a number of machines and
>>reorganize to a RAID-backed fileserver.
>>
>>The machine is a dual pentium-pro 200 (320MB RAM) .. would that be a
>>dumb idea to use RAID5 on it because of the parity calculations
>>needed?
>>
>>Further to that, would it be a smarter idea to use RAID1 on all 4 of
>>some small partition(s) at the start of the disks to house
>>boot/root/usr partitions, and only RAID5 on a larger 'data' area of
>>the drive that is more likely to be read than written to?

Derek,

I have a machine with 6 x 250GB SATA disks, but the configuration I use 
would work just as well for you. Here's what I'd do:

Partition all your drives the same.
Create one small partition of 1GB, plus one large partition using up the 
rest of the disk (i.e. around 59GB), *except* the 80GB drive. On this, 
create a 1GB partition, a 59GB partition, plus a third partition using 
up the rest of the disk (i.e. around 20GB)

Assuming these drives are /dev/hd[efgh], configure them as follows:

/dev/hd[ef]1  	/dev/md0	/
/dev/hd[gh]1	/dev/md1	swap
/dev/hd[efgh]2	/dev/md2	lvm volume group
/dev/hdh3	-		use for whatever you want!

Now, use lvm to create logical volumes in your large volume group. I 
have created /var, /use, and use the rest for /home.

These are my arrays:

[root@dude slimserver]# mdadm --detail --scan
ARRAY /dev/md1 level=raid1 num-devices=2 
UUID=be8ad31a:f13b6f4b:c39732fc:c84f32a8
    devices=/dev/sdb1,/dev/sde1
ARRAY /dev/md2 level=raid1 num-devices=2 
UUID=826170e2:cdd598d4:d212c9b1:6602deef
    devices=/dev/sdc1,/dev/sdf1
ARRAY /dev/md5 level=raid5 num-devices=5 spares=1 
UUID=a4bbcd09:5e178c5b:3bf8bd45:8c31d2a1
    devices=/dev/sda2,/dev/sdb2,/dev/sdc2,/dev/sdd2,/dev/sde2,/dev/sdf2
ARRAY /dev/md0 level=raid1 num-devices=2 
UUID=4b28338c:bf08d0bc:bb2899fc:e7f35eae
    devices=/dev/sda1,/dev/sdd1

These are the lvm logical volumes:

[root@dude slimserver]# lvdisplay
   --- Logical volume ---
   LV Name                /dev/audio_vg/usr_lv
   VG Name                audio_vg
   LV UUID                qseH0A-wKgo-xhB5-2tJ4-Qnxx-VOML-0eb43m
   LV Write Access        read/write
   LV Status              available
   # open                 1
   LV Size                10.00 GB
   Current LE             160
   Segments               1
   Allocation             inherit
   Read ahead sectors     0
   Block device           253:0

   --- Logical volume ---
   LV Name                /dev/audio_vg/var_lv
   VG Name                audio_vg
   LV UUID                nzH8uf-LhyU-o5My-tK48-ckaw-xzfL-esbfj4
   LV Write Access        read/write
   LV Status              available
   # open                 1
   LV Size                5.00 GB
   Current LE             80
   Segments               1
   Allocation             inherit
   Read ahead sectors     0
   Block device           253:1

   --- Logical volume ---
   LV Name                /dev/audio_vg/home_lv
   VG Name                audio_vg
   LV UUID                zbixtc-S6mb-MTVR-WXGw-dkjG-EU9q-WeZItv
   LV Write Access        read/write
   LV Status              available
   # open                 1
   LV Size                914.38 GB
   Current LE             14630
   Segments               1
   Allocation             inherit
   Read ahead sectors     0
   Block device           253:2


This is what my filesystems look like:

[root@dude slimserver]# df -h
Filesystem            Size  Used Avail Use% Mounted on
/dev/md0              1.4G  357M  985M  27% /
/dev/mapper/audio_vg-var_lv
                       5.0G  1.4G  3.3G  30% /var
/dev/mapper/audio_vg-usr_lv
                       9.9G  2.4G  7.0G  26% /usr
/dev/mapper/audio_vg-home_lv
                       915G  142G  764G  16% /home

And finally swap:

[root@dude slimserver]# swapon -s
Filename                                Type            Size    Used 
Priority
/dev/md1                                partition       1469816 224     -1

R.
--
http://robinbowes.com

Re: RAID5 on different sized disks on low-end machine

[Derek Piper]

Thanks for that information, that sounds like a good idea. The only
thing that concerns me is that in googling for LVM, I'm worried that
it's an extra layer of complication that may not be that stable as yet
since it needs devfs .. or am I talking out of my a$$ and have seen
too many scare stories? I like your idea of splitting the swap and
root partitions out to separate RAID1s.. I guess I could do something
similar even without LVM.

Thanks :)

Derek

/paranoid he's gonna choose the wrong thing and hose himself 6months
down the line.

On Wed, 12 Jan 2005 00:26:33 +0000, Robin Bowes
<robin-lists@robinbowes.com> wrote:
> Derek Piper wrote:
> > Hi,
> >
> > I revised my idea and thought about RAID 1+0 for some partitions,
> > since there are 4 drives. This outline below might clarify what I was
> > trying to mention earlier. Is this a feasible set-up that would be
> > bootable (kernel compiled-in md, I'm no stranger to compiling
> > kernels)? I'm interested to hear comments/opinions since I've never
> > done this before. Like I said, it'll be running on a Dual-pentium pro
> > 200 (W6-LI) machine, I have no idea if machines of that vintage have
> > the 'cojones' for software raid or not.
> >
> > My ideas of RAID1+0 / RAID5 disk system partitions
> >               MB
> > /dev/hde      60GB    57241   (from controller)
> > /dev/hdf      60GB    57241   (from controller)
> > /dev/hdg      60GB    57241   (from controller)
> > /dev/hdh      80GB    78125   (unconfirmed)
> >
> > /dev/hd* = applies to all drives considered here
> >
> > Device        MB      Type    GB      Mountpoint      MD device       RAIDed size (MB)        GB
> > /dev/hd*1     20      RAID1 + 0       0.02    /boot   /dev/md1        40      0.04
> > /dev/hd*2     192     RAID1 + 0       0.19    Swap    /dev/md2        384     0.38
> > /dev/hd*5     2048    RAID1 + 0       2       /       /dev/md5        4096    4
> > /dev/hd*6     2048    RAID5   2       /home   /dev/md6        6144    6
> > /dev/hd*7     52933   RAID5   51.69   /data   /dev/md7        158799  155.08
> >
> > Does swap being raided make sense? I hear that sometimes it's a good
> > idea since a disk failure won't make you crash and then I heard
> > elsewhere that it doesn't matter and the kernel automatically raids
> > swap partitions anyway. I prepared the above in a spreadsheet btw so I
> > could work out partition sizes.
> >
> > Thanks in advance again for any comments.
> >
> > Derek
> >
> > On Tue, 11 Jan 2005 13:47:20 -0500, Derek Piper <derek.piper@gmail.com> wrote:
> >
> >>Hi,
> >>
> >>I am new to RAID / md devices, although I've used Linux for a number
> >>of years. I decided it was high-time I had a RAID at home for
> >>important things (email, web-sites, son's baby pics, mp3s etc.). I
> >>happen to have a 3 Seagate 60GB hds and 1 80GB Seagate hd that I am
> >>considering using for a RAID.
> >>
> >>My question is this, is it possible (and even a good idea) to use all
> >>4 hard drives as members of a 4 x 60GB RAID5 array by leaving 20GB of
> >>the 80GB drive as a non-raided partition? I'll be using a Promise
> >>Ultra TX2/100 controller.
> >>
> >>i.e.
> >>
> >>hde -> 60
> >>hdf -> 60
> >>hdg -> 60
> >>hdh -> 60/20
> >>
> >>I heard about RAID6 too, though I'm assuming that will use up another
> >>disk's worth of disk space too.
> >>
> >>i.e. RAID5 = 180GB usable size,wherease RAID6 = 120GB .. am I correct
> >>in my thinking?
> >>
> >>I know many of you use far larger hard drives, I'm just trying to use
> >>the components I already had spare from a number of machines and
> >>reorganize to a RAID-backed fileserver.
> >>
> >>The machine is a dual pentium-pro 200 (320MB RAM) .. would that be a
> >>dumb idea to use RAID5 on it because of the parity calculations
> >>needed?
> >>
> >>Further to that, would it be a smarter idea to use RAID1 on all 4 of
> >>some small partition(s) at the start of the disks to house
> >>boot/root/usr partitions, and only RAID5 on a larger 'data' area of
> >>the drive that is more likely to be read than written to?
> 
> Derek,
> 
> I have a machine with 6 x 250GB SATA disks, but the configuration I use
> would work just as well for you. Here's what I'd do:
> 
> Partition all your drives the same.
> Create one small partition of 1GB, plus one large partition using up the
> rest of the disk (i.e. around 59GB), *except* the 80GB drive. On this,
> create a 1GB partition, a 59GB partition, plus a third partition using
> up the rest of the disk (i.e. around 20GB)
> 
> Assuming these drives are /dev/hd[efgh], configure them as follows:
> 
> /dev/hd[ef]1    /dev/md0        /
> /dev/hd[gh]1    /dev/md1        swap
> /dev/hd[efgh]2  /dev/md2        lvm volume group
> /dev/hdh3       -               use for whatever you want!
> 
> Now, use lvm to create logical volumes in your large volume group. I
> have created /var, /use, and use the rest for /home.
> 
> These are my arrays:
> 
> [root@dude slimserver]# mdadm --detail --scan
> ARRAY /dev/md1 level=raid1 num-devices=2
> UUID=be8ad31a:f13b6f4b:c39732fc:c84f32a8
>     devices=/dev/sdb1,/dev/sde1
> ARRAY /dev/md2 level=raid1 num-devices=2
> UUID=826170e2:cdd598d4:d212c9b1:6602deef
>     devices=/dev/sdc1,/dev/sdf1
> ARRAY /dev/md5 level=raid5 num-devices=5 spares=1
> UUID=a4bbcd09:5e178c5b:3bf8bd45:8c31d2a1
>     devices=/dev/sda2,/dev/sdb2,/dev/sdc2,/dev/sdd2,/dev/sde2,/dev/sdf2
> ARRAY /dev/md0 level=raid1 num-devices=2
> UUID=4b28338c:bf08d0bc:bb2899fc:e7f35eae
>     devices=/dev/sda1,/dev/sdd1
> 
> These are the lvm logical volumes:
> 
> [root@dude slimserver]# lvdisplay
>    --- Logical volume ---
>    LV Name                /dev/audio_vg/usr_lv
>    VG Name                audio_vg
>    LV UUID                qseH0A-wKgo-xhB5-2tJ4-Qnxx-VOML-0eb43m
>    LV Write Access        read/write
>    LV Status              available
>    # open                 1
>    LV Size                10.00 GB
>    Current LE             160
>    Segments               1
>    Allocation             inherit
>    Read ahead sectors     0
>    Block device           253:0
> 
>    --- Logical volume ---
>    LV Name                /dev/audio_vg/var_lv
>    VG Name                audio_vg
>    LV UUID                nzH8uf-LhyU-o5My-tK48-ckaw-xzfL-esbfj4
>    LV Write Access        read/write
>    LV Status              available
>    # open                 1
>    LV Size                5.00 GB
>    Current LE             80
>    Segments               1
>    Allocation             inherit
>    Read ahead sectors     0
>    Block device           253:1
> 
>    --- Logical volume ---
>    LV Name                /dev/audio_vg/home_lv
>    VG Name                audio_vg
>    LV UUID                zbixtc-S6mb-MTVR-WXGw-dkjG-EU9q-WeZItv
>    LV Write Access        read/write
>    LV Status              available
>    # open                 1
>    LV Size                914.38 GB
>    Current LE             14630
>    Segments               1
>    Allocation             inherit
>    Read ahead sectors     0
>    Block device           253:2
> 
> This is what my filesystems look like:
> 
> [root@dude slimserver]# df -h
> Filesystem            Size  Used Avail Use% Mounted on
> /dev/md0              1.4G  357M  985M  27% /
> /dev/mapper/audio_vg-var_lv
>                        5.0G  1.4G  3.3G  30% /var
> /dev/mapper/audio_vg-usr_lv
>                        9.9G  2.4G  7.0G  26% /usr
> /dev/mapper/audio_vg-home_lv
>                        915G  142G  764G  16% /home
> 
> And finally swap:
> 
> [root@dude slimserver]# swapon -s
> Filename                                Type            Size    Used
> Priority
> /dev/md1                                partition       1469816 224     -1
> 
> R.
> --
> http://robinbowes.com
> 
> -
> To unsubscribe from this list: send the line "unsubscribe linux-raid" in
> the body of a message to majordomo@vger.kernel.org
> More majordomo info at  http://vger.kernel.org/majordomo-info.html
> 


-- 
Derek Piper - derek.piper@gmail.com
http://doofer.org/

Re: RAID5 on different sized disks on low-end machine

[Norman Schmidt]

Hi Derek!

What worries me a little bit about your setup is that you seem to want 
to put the drives as master and slave on each of the two busses of the 
ata controller. I think this is not so good an idea, because on one 
hand, it is slower (I tested this with a raid1 on two disks, 
master/slave on one and master/master on two channels of a Promise Ultra 
133 TX2 under 2.4), and on the other, it is not so secure. The reason is 
that usually one disk going down (or one faulty cable) on an ide bus 
takes down the other drive on it as well so you would lose two of four 
(which is fatal for raid5), and would have to hope that you can 
force-resync the raid after unplugging the faulty disk.
It is a bit of luxury (and takes more pci slots and will perhaps produce 
a little more overhead on the pci bus), but I would use two controllers 
with four channels and only master disks (that´s what I am doing in two 
cases at the moment). What I also found out is that more than two of the 
exact same controllers on the pci bus don´t work (under 2.4), so in the 
two aforementioned cases I use two Promise U133 and one U100 (5 and 6 
disks respectively).

Norman.



-- 
Norman Schmidt          Institut fuer Physikal. u. Theoret. Chemie
Dipl.-Chem. Univ.       Friedrich-Alexander-Universitaet
schmidt@naa.net         Erlangen-Nuernberg
                         IT-Systembetreuer Physikalische Chemie
-
To unsubscribe from this list: send the line "unsubscribe linux-raid" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html

Re: RAID5 on different sized disks on low-end machine

[Robin Bowes]

Norman Schmidt wrote:
> Hi Derek!
> 
> What worries me a little bit about your setup is that you seem to want 
> to put the drives as master and slave on each of the two busses of the 
> ata controller.

Ah, I missed that. Yes, that's definitely not a good idea. Get yourself 
another controller card and stick to one disk per ATA channel.

R.
-- 
http://robinbowes.com

Fwd: RAID5 on different sized disks on low-end machine

[Derek Piper]

Hi all,

Yes, I did notice that I would be doing the master/slave thing and
went and read to the FAQ about it (omg! he reads?!). I actually set up
(well, am in the process of setting up) the machine to use it's older
ATA33 ports of hda and hdc. I was going to post my revision of idea to
the list but forgot. Thanks for responding to it though! :)

I would get another promise card, but since I need two NICs in this
machine (Dual PPro W6LI board) and the VGA card (PCI), all PCI slots
are used.

I went with this layout in the end:

/dev/hde,/dev/hdg,/dev/hdc = 60000MB
/dev/hda = 80000MB
Pair1 = /dev/hde, /dev/hdg on the Promise Ultra TX2 (bootable, ATA100,
boots to large drives)
Pair2 = /dev/hda, /dev/hdc on the motherboard IDE controllers (ATA33
and BIOS cannot boot to large drives)

/dev/hd*1       32      RAID1   0.03    /boot   /dev/md1        32    
 0.03    Pair1
/dev/hd*2       768     RAID1   0.75    Swap    /dev/md0        768   
 0.75    Pair1
/dev/hd*5       1760    RAID1   1.72    /       /dev/md2        1760  
 1.72    Pair1

/dev/hd*5       2560    RAID1   2.5     /usr    /dev/md3        2560  
 2.5     Pair2

/dev/hd*6       57462   RAID5   56.12   /mnt/array      /dev/md4      
 172386  168.35  (all)
/dev/hda7       20663   (non-raid)      20.18

One thing though, since I'm posting to the list...

I have the machine able to boot from RAID, and the filesystem is
copied to the above assigned md devices. I added the initial
installation drive (happened to be /dev/hde) and to the md4 array to
complete the RAID5. All RAID5 component partitions are the exact same
block size according to fdisk -l , so at least partitioning the larger
disk was okay.

The problem is now that during the course of the re-sync (it added hde
in as 'spare', is that normal?) it tells me that /dev/hdc has 'failed'
and decides to kick it out. That's a bit worrysome, since what happens
if there was data on that array? Does that tend to happen? Is RAID5
REALLY that unreliable? Jokes about p0rn and replacable data aside,
should we really be trusting valuable data to it? What are the
opinions out there?

I'm wondering about just saying 'screw it' to RAID5 and doing two
RAID1 mirrors with the partitions instead since it's data I really do
not want to lose that would be going on there. I've never had any
problems with any of the HDs, even hdc, so it's quite surprising it
would barf like that. The kernel didn't give any messages and I've
since rebooted. I was tinkering with it from work during my lunch
break, getting the RAID arrays created and booting from RAID, but now
after rebooting it's not come back so that I can log into it again. I
guess I'll see what it's complaining about this evening :>

Derek

On Fri, 14 Jan 2005 10:07:03 +0000, Robin Bowes
<robin-lists@robinbowes.com> wrote:
> Norman Schmidt wrote:
> > Hi Derek!
> >
> > What worries me a little bit about your setup is that you seem to want
> > to put the drives as master and slave on each of the two busses of the
> > ata controller.
>
> Ah, I missed that. Yes, that's definitely not a good idea. Get yourself
> another controller card and stick to one disk per ATA channel.
>
> R.
> --
> http://robinbowes.com
>
> -
> To unsubscribe from this list: send the line "unsubscribe linux-raid" in
> the body of a message to majordomo@vger.kernel.org
> More majordomo info at  http://vger.kernel.org/majordomo-info.html
>

-- 
Derek Piper - derek.piper@gmail.com
http://doofer.org/

Re: Fwd: RAID5 on different sized disks on low-end machine

[berk walker]

If you have an AGP slot, dump the pci vga, eh?
b-

Derek Piper wrote:

>Hi all,
>
>Yes, I did notice that I would be doing the master/slave thing and
>went and read to the FAQ about it (omg! he reads?!). I actually set up
>(well, am in the process of setting up) the machine to use it's older
>ATA33 ports of hda and hdc. I was going to post my revision of idea to
>the list but forgot. Thanks for responding to it though! :)
>
>I would get another promise card, but since I need two NICs in this
>machine (Dual PPro W6LI board) and the VGA card (PCI), all PCI slots
>are used.
>
>I went with this layout in the end:
>
>/dev/hde,/dev/hdg,/dev/hdc = 60000MB
>/dev/hda = 80000MB
>Pair1 = /dev/hde, /dev/hdg on the Promise Ultra TX2 (bootable, ATA100,
>boots to large drives)
>Pair2 = /dev/hda, /dev/hdc on the motherboard IDE controllers (ATA33
>and BIOS cannot boot to large drives)
>
>/dev/hd*1       32      RAID1   0.03    /boot   /dev/md1        32    
> 0.03    Pair1
>/dev/hd*2       768     RAID1   0.75    Swap    /dev/md0        768   
> 0.75    Pair1
>/dev/hd*5       1760    RAID1   1.72    /       /dev/md2        1760  
> 1.72    Pair1
>
>/dev/hd*5       2560    RAID1   2.5     /usr    /dev/md3        2560  
> 2.5     Pair2
>
>/dev/hd*6       57462   RAID5   56.12   /mnt/array      /dev/md4      
> 172386  168.35  (all)
>/dev/hda7       20663   (non-raid)      20.18
>
>One thing though, since I'm posting to the list...
>
>I have the machine able to boot from RAID, and the filesystem is
>copied to the above assigned md devices. I added the initial
>installation drive (happened to be /dev/hde) and to the md4 array to
>complete the RAID5. All RAID5 component partitions are the exact same
>block size according to fdisk -l , so at least partitioning the larger
>disk was okay.
>
>The problem is now that during the course of the re-sync (it added hde
>in as 'spare', is that normal?) it tells me that /dev/hdc has 'failed'
>and decides to kick it out. That's a bit worrysome, since what happens
>if there was data on that array? Does that tend to happen? Is RAID5
>REALLY that unreliable? Jokes about p0rn and replacable data aside,
>should we really be trusting valuable data to it? What are the
>opinions out there?
>
>I'm wondering about just saying 'screw it' to RAID5 and doing two
>RAID1 mirrors with the partitions instead since it's data I really do
>not want to lose that would be going on there. I've never had any
>problems with any of the HDs, even hdc, so it's quite surprising it
>would barf like that. The kernel didn't give any messages and I've
>since rebooted. I was tinkering with it from work during my lunch
>break, getting the RAID arrays created and booting from RAID, but now
>after rebooting it's not come back so that I can log into it again. I
>guess I'll see what it's complaining about this evening :>
>
>Derek
>
>On Fri, 14 Jan 2005 10:07:03 +0000, Robin Bowes
><robin-lists@robinbowes.com> wrote:
>  
>
>>Norman Schmidt wrote:
>>    
>>
>>>Hi Derek!
>>>
>>>What worries me a little bit about your setup is that you seem to want
>>>to put the drives as master and slave on each of the two busses of the
>>>ata controller.
>>>      
>>>
>>Ah, I missed that. Yes, that's definitely not a good idea. Get yourself
>>another controller card and stick to one disk per ATA channel.
>>
>>R.
>>--
>>http://robinbowes.com
>>
>>-
>>To unsubscribe from this list: send the line "unsubscribe linux-raid" in
>>the body of a message to majordomo@vger.kernel.org
>>More majordomo info at  http://vger.kernel.org/majordomo-info.html
>>
>>    
>>
>
>  
>