Preparation advice?

Preparation advice?

[Ewan Grantham]

Am about to venture into the MD world. I have 4 250-Gig HDs, 1 on
IDE1, 1 on IDE2, 1 on the internal firewire port, 1 on a PCI Firewire
card. So I think I have my failure points reasonably spread :-)

Since two of these are external drives, any special considerations? I
am running a Debian from a Knoppix 3.6 HD Install with a 2.4.27
kernel. I just updated the mdadm package using synaptic to 1.7, which
I assume is the best version to run. My OS/Boot Disk is seperate, so I
don't have to worry about that, but would like to have the array come
up when the system does. Pointers to FAQs on any of this would also be
appreciated.

Thanks in advance,
Ewan

Re: Preparation advice?

[Gordon Henderson]

On Fri, 26 Nov 2004, Ewan Grantham wrote:

> Am about to venture into the MD world. I have 4 250-Gig HDs, 1 on
> IDE1, 1 on IDE2, 1 on the internal firewire port, 1 on a PCI Firewire
> card. So I think I have my failure points reasonably spread :-)

Sounds like a bit of a homebrew bits & pieces box... :)

> Since two of these are external drives, any special considerations?

Make sure their power and data cables, and on/off switches are secure...

> I
> am running a Debian from a Knoppix 3.6 HD Install with a 2.4.27
> kernel. I just updated the mdadm package using synaptic to 1.7, which
> I assume is the best version to run. My OS/Boot Disk is seperate, so I
> don't have to worry about that, but would like to have the array come
> up when the system does. Pointers to FAQs on any of this would also be
> appreciated.

Read the Software RAID FAQ on

  http://www.tldp.org/HOWTO/Software-RAID-HOWTO.html

It's got basic instructions for both the older raidtools and the newer
mdadm.

What is your boot disk connected to? Hopefully not the same controller as
one of your RAID disks...

And might you not want to make the bood drive mirrored also?

Anyway, for the big array, assuming you just want one big ~750GB array
(RAID 5 gives N-1 capacity) partition each drive the same way with one
partition for the whole disk, then mark each partition with type FD and
build the array. Make sure the kernel has RAID built-in to it, or the RAID
modules are loaded at boot time and off you go. Once it's built, make a
mount point for it and put an entry in the /etc/fstab for it and it'll be
there for all time....

Gordon

Re: Preparation advice?

[Ewan Grantham]

On Sat, 27 Nov 2004 08:46:37 +0000 (GMT), Gordon Henderson
<gordon@drogon.net> wrote:
> Sounds like a bit of a homebrew bits & pieces box... :)

Well, it started out as a Compaq Presario, but as time has gone on...

> Read the Software RAID FAQ on
> 
>   http://www.tldp.org/HOWTO/Software-RAID-HOWTO.html
> 
> It's got basic instructions for both the older raidtools and the newer
> mdadm.

Is that going to ever be updated? I saw that site, as well as the
helpful chapter extract for using MDADM on O'Reilly, and noticed both
were from about 2002. I understand that the tools still work basically
the same way, but wonder if "our" understanding of RAID hasn't come
along a bit since then. And there are other topics that would be nice
to address - that I will be asking about shortly :-)

> And might you not want to make the bood drive mirrored also?

My boot drive is OS only. Idea has been (from past experience) to not
have anything on the boot drive that I wouldn't mind wiping for a
reinstall.

> Anyway, for the big array, assuming you just want one big ~750GB array
> (RAID 5 gives N-1 capacity) partition each drive the same way with one
> partition for the whole disk, then mark each partition with type FD and
> build the array. Make sure the kernel has RAID built-in to it, or the RAID
> modules are loaded at boot time and off you go. Once it's built, make a
> mount point for it and put an entry in the /etc/fstab for it and it'll be
> there for all time....

Currently building md0 as RAID 5 with chunk of 128 since all the
examples seem to indicate that is "better" than the 64 default.

Created a mount point /mnt/md0, but do I edit /etc/fstab or another
file since fstab seems to be generated on boot?

Also, once the build of md0 is complete, I assume I need to format my
new "drive". Any suggestions as to ext2 or ext3 or reiser or ...

Array will be used to capture and edit home videos - to the extent
that would influence folks suggestions.

Thanks again,
Ewan

Re: Preparation advice?

[Ewan Grantham]

On Sat, 27 Nov 2004 18:20:42 +0000, David Greaves <david@dgreaves.com> wrote:
> 
> >Created a mount point /mnt/md0, but do I edit /etc/fstab or another
> >file since fstab seems to be generated on boot?
> >
> >
> edit /etc/fstab

Well, I must be missing this at the FAQ sites, but when I edit
/etc/fstab do I leave the entries in for the component devices, or do
I remove those and just put in the /dev/md0 entry?

Thanks again,
Ewan

Re: Preparation advice?

[David Greaves]

Well, there's nothing wrong with /etc/fstab etc but you do have an odd 
report from mdadm.
Back to the list... :)

Ewan's running 2.4.27 and may have an inconsistency with mdadm counting 
the number of raid devices.



Ewan Grantham wrote:

> also send me the output from:
>  mdadm --detail /dev/md0

/dev/md0:
        Version : 00.90.00
  Creation Time : Sat Nov 27 07:32:34 2004
     Raid Level : raid5
     Array Size : 735334656 (701.27 GiB 752.98 GB)
    Device Size : 245111552 (233.76 GiB 250.99 GB)
   Raid Devices : 4
  Total Devices : 5
Preferred Minor : 0
    Persistence : Superblock is persistent

    Update Time : Sat Nov 27 14:14:24 2004
          State : dirty
 Active Devices : 4
Working Devices : 4
 Failed Devices : 1
  Spare Devices : 0

         Layout : left-symmetric
     Chunk Size : 128K

           UUID : 525b6c6e:836b8598:b3778337:0be081e8
         Events : 0.2

    Number   Major   Minor   RaidDevice State
       0       8       33        0      active sync   /dev/sdc1
       1       8       49        1      active sync   /dev/sdd1
       2       3       65        2      active sync   /dev/hdb1
       3      22       65        3      active sync   /dev/hdd1



>>I wonder if you have an older version of mdadm?
>>mdadm -V
>>can you upgrade to 1.7 ?
>>    
>>
>
>mdadm - v1.7.0 - 11 August 2004
>
>  
>
>>What command did you use to create the array?
>>    
>>
>
>mdadm --create --verbose /dev/md0 --level=5 -c128 
>        --raid-devices=4  /dev/sdc1 /dev/sdd1 /dev/hdb1 /dev/hdd1
>
>  
>
>>What kernel version are you running?
>>    
>>
>
>2.4.27
>
>  
>
>>Have you put data on the array yet? (if not we may blat and rebuild it)
>>    
>>
>
>Yes, but nothing I can't live without at this point. More a matter of
>the time it takes - about 6 hours. But would rather take another 6
>hours than be likely to lose what I have on there...
>
>Let me know if there's something else I should look at/do...
>  
>
what does
 cat /proc/mdstat
say?


David

Re: Preparation advice?

[Ewan Grantham]

On Sun, 28 Nov 2004 11:04:51 +0000, David Greaves <david@dgreaves.com> wrote:
> Ewan's running 2.4.27 and may have an inconsistency with mdadm counting
> the number of raid devices.
...
> /dev/md0:
>         Version : 00.90.00
>   Creation Time : Sat Nov 27 07:32:34 2004
>      Raid Level : raid5
>      Array Size : 735334656 (701.27 GiB 752.98 GB)
>     Device Size : 245111552 (233.76 GiB 250.99 GB)
>    Raid Devices : 4
>   Total Devices : 5
> Preferred Minor : 0
>     Persistence : Superblock is persistent
> 
>     Update Time : Sat Nov 27 14:14:24 2004
>           State : dirty
>  Active Devices : 4
> Working Devices : 4
>  Failed Devices : 1
>   Spare Devices : 0
...
> what does
>  cat /proc/mdstat
> say?

As noted above, I'm getting an "interesting" discrepancy between the 4
devices I specified in my create and the results from mdadm. The
device seems to be working fine after transferring and playing several
files, but I find the "failed devices" bit particularly concerning.
Haven't found any way to get mdadm to tell me which device it thinks
has failed.

As for the command above, I get:
Personalities : [raid5]
read_ahead 1024 sectors
md0 : active raid5 hdd1[3] hdb1[2] sdd1[1] sdc1[0]
      735334656 blocks level 5, 128k chunk, algorithm 2 [4/4] [UUUU]

unused devices: <none>

Which is what I would have expected. Any ideas?

RE: Preparation advice?

[Guy]

This is normal (IMO) for a 2.4 kernel.
I think it has been fixed in the 2.6 kernel.  But I have never used the
newer kernel, so I can't confirm that.  It may have been a newer version of
mdadm, not the kernel, not sure.

My numbers are much worse!
I have 14 disks and 1 spare.
   Raid Devices : 14
  Total Devices : 13

 Active Devices : 14
Working Devices : 12
 Failed Devices : 1
  Spare Devices : 1

    Number   Major   Minor   RaidDevice State
       0       8       49        0      active sync   /dev/sdd1
       1       8      161        1      active sync   /dev/sdk1
       2       8       65        2      active sync   /dev/sde1
       3       8      177        3      active sync   /dev/sdl1
       4       8       81        4      active sync   /dev/sdf1
       5       8      193        5      active sync   /dev/sdm1
       6       8       97        6      active sync   /dev/sdg1
       7       8      209        7      active sync   /dev/sdn1
       8       8      113        8      active sync   /dev/sdh1
       9       8      225        9      active sync   /dev/sdo1
      10       8      129       10      active sync   /dev/sdi1
      11       8      241       11      active sync   /dev/sdp1
      12       8      145       12      active sync   /dev/sdj1
      13       8       33       13      active sync   /dev/sdc1
      14      65        1       14        /dev/sdq1

Guy

-----Original Message-----
From: linux-raid-owner@vger.kernel.org
[mailto:linux-raid-owner@vger.kernel.org] On Behalf Of Ewan Grantham
Sent: Sunday, November 28, 2004 8:37 AM
To: David Greaves
Cc: linux-raid@vger.kernel.org
Subject: Re: Preparation advice?

On Sun, 28 Nov 2004 11:04:51 +0000, David Greaves <david@dgreaves.com>
wrote:
> Ewan's running 2.4.27 and may have an inconsistency with mdadm counting
> the number of raid devices.
...
> /dev/md0:
>         Version : 00.90.00
>   Creation Time : Sat Nov 27 07:32:34 2004
>      Raid Level : raid5
>      Array Size : 735334656 (701.27 GiB 752.98 GB)
>     Device Size : 245111552 (233.76 GiB 250.99 GB)
>    Raid Devices : 4
>   Total Devices : 5
> Preferred Minor : 0
>     Persistence : Superblock is persistent
> 
>     Update Time : Sat Nov 27 14:14:24 2004
>           State : dirty
>  Active Devices : 4
> Working Devices : 4
>  Failed Devices : 1
>   Spare Devices : 0
...
> what does
>  cat /proc/mdstat
> say?

As noted above, I'm getting an "interesting" discrepancy between the 4
devices I specified in my create and the results from mdadm. The
device seems to be working fine after transferring and playing several
files, but I find the "failed devices" bit particularly concerning.
Haven't found any way to get mdadm to tell me which device it thinks
has failed.

As for the command above, I get:
Personalities : [raid5]
read_ahead 1024 sectors
md0 : active raid5 hdd1[3] hdb1[2] sdd1[1] sdc1[0]
      735334656 blocks level 5, 128k chunk, algorithm 2 [4/4] [UUUU]

unused devices: <none>

Which is what I would have expected. Any ideas?
-
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: Preparation advice?

[Ewan Grantham]

On Sun, 28 Nov 2004 13:00:54 -0500, Guy <bugzilla@watkins-home.com> wrote:
> This is normal (IMO) for a 2.4 kernel.
> I think it has been fixed in the 2.6 kernel.  But I have never used the
> newer kernel, so I can't confirm that.  It may have been a newer version of
> mdadm, not the kernel, not sure.

A rather different definition of normal - but for a computer system I
guess I shouldn't be that surprised :-)

I'm not overly inclined to move to the 2.6 kernel at the moment as my
experience with IEEE1394 under 2.6 has been very mixed. I think it's
because it's easier to see what's happening "under the hood" with the
2.4 code.

Since I gather you haven't seen any lost data from this, then I'm
going to start filling up. Did a couple test captures today without
any dropped frames, so I'm feeling a bit more comfortable that this is
going to work.

Assuming I bought some more drives in the future, would I be better
off getting another firewire card or two, or going with USB 2?

Thanks again for all the help and advice,
Ewan

RE: Preparation advice?

[Guy]

I have not had any corruption or any other problems, other than when a disk
fails.  Even then, never had data loss.  I just ignore the bogus numbers.

Guy

-----Original Message-----
From: linux-raid-owner@vger.kernel.org
[mailto:linux-raid-owner@vger.kernel.org] On Behalf Of Ewan Grantham
Sent: Sunday, November 28, 2004 6:33 PM
To: Guy
Cc: David Greaves; linux-raid@vger.kernel.org
Subject: Re: Preparation advice?

On Sun, 28 Nov 2004 13:00:54 -0500, Guy <bugzilla@watkins-home.com> wrote:
> This is normal (IMO) for a 2.4 kernel.
> I think it has been fixed in the 2.6 kernel.  But I have never used the
> newer kernel, so I can't confirm that.  It may have been a newer version
of
> mdadm, not the kernel, not sure.

A rather different definition of normal - but for a computer system I
guess I shouldn't be that surprised :-)

I'm not overly inclined to move to the 2.6 kernel at the moment as my
experience with IEEE1394 under 2.6 has been very mixed. I think it's
because it's easier to see what's happening "under the hood" with the
2.4 code.

Since I gather you haven't seen any lost data from this, then I'm
going to start filling up. Did a couple test captures today without
any dropped frames, so I'm feeling a bit more comfortable that this is
going to work.

Assuming I bought some more drives in the future, would I be better
off getting another firewire card or two, or going with USB 2?

Thanks again for all the help and advice,
Ewan
-
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: Preparation advice?

[Gordon Henderson]

On Sat, 27 Nov 2004, Ewan Grantham wrote:

> On Sat, 27 Nov 2004 08:46:37 +0000 (GMT), Gordon Henderson
> <gordon@drogon.net> wrote:
> > Sounds like a bit of a homebrew bits & pieces box... :)
>
> Well, it started out as a Compaq Presario, but as time has gone on...
>
> > Read the Software RAID FAQ on
> >
> >   http://www.tldp.org/HOWTO/Software-RAID-HOWTO.html
> >
> > It's got basic instructions for both the older raidtools and the newer
> > mdadm.
>
> Is that going to ever be updated? I saw that site, as well as the
> helpful chapter extract for using MDADM on O'Reilly, and noticed both
> were from about 2002. I understand that the tools still work basically
> the same way, but wonder if "our" understanding of RAID hasn't come
> along a bit since then. And there are other topics that would be nice
> to address - that I will be asking about shortly :-)

Er, I'm sure if you want to volenteer to update it your thoughts will be
taken on-board :)

> > And might you not want to make the bood drive mirrored also?
>
> My boot drive is OS only. Idea has been (from past experience) to not
> have anything on the boot drive that I wouldn't mind wiping for a
> reinstall.

Fair-enough, but with 4 big disks, taking a GB to put the OS on, and have
it RAID1'd over the 4 disks might just save you that half-days agony if
the boot disk ever fails..

> Currently building md0 as RAID 5 with chunk of 128 since all the
> examples seem to indicate that is "better" than the 64 default.

"Better" really depends on what you might be using it for - but it sounds
like it might be OK if all you are doing is streaming large files... The
only realy way to work out is to benchmark the array with something thats
as close to your application as possible.

Remember that you can reflect the chunk size in the ext2/3 strip size...
(And maybe for other filesystems too)

> Created a mount point /mnt/md0, but do I edit /etc/fstab or another
> file since fstab seems to be generated on boot?

I don't know any distros that denetate this at boot time, but I only
really know Debian and a bit of deadrat these days. Traditionally you
would edit this file to add in your array under some mount point of your
choosing.

> Also, once the build of md0 is complete, I assume I need to format my
> new "drive". Any suggestions as to ext2 or ext3 or reiser or ...

You can format is while it's building. However, in-light of the recent
threads on drive failures, etc. it's well worth your time to run badblocks
on it overnight with both read and write tests just to make sure that
every block can be actually written and read back again, although just
waiting for the raid5 to sync will do some of this, but not everything. If
you have the time, I'd recomend you do it.

Also make sure you have the right drivers loaded in the kernel - it's
worth while doing a gross speed test on each device using hdparm to check
this - eg. hdparm -Tt /dev/sad1 etc.

I'd build the array, wait for it to sync, run badblocks -sw -p99 -c8192 on
it overnight, then mkfs it with your favouring filesystem.

Holy wars have been fought over lesser issues than filesystem choice, so
you might as well toss a coin, or stick to something you know and like. A
journalling filesystem (eg. ext3, xfs, reiser) may impose some latency on
long streaming writes which is what you are after, but only benchmarking
will tell you this.

Personally, I'd stick to ext2 or 3, having used it for many years now, but
I might have a look at XFS again, which I looked at some 12-18 months back
but didn't really have good results with it.

> Array will be used to capture and edit home videos - to the extent
> that would influence folks suggestions.

Enjoy!

Gordon