FAQ Search Today's Posts Mark Forums Read
» Video Reviews

» Linux Archive

Linux-archive is a website aiming to archive linux email lists and to make them easily accessible for linux users/developers.


» Sponsor

» Partners

» Sponsor

Go Back   Linux Archive > Debian > Debian User

 
 
LinkBack Thread Tools
 
Old 03-11-2009, 06:52 AM
Clinton Lee Taylor
 
Default Mirror RAID convert ...

Greetings ...

Have seen similar dicsuss here before, but can't seem to find the
messages now ...

Was thinking or hoping, that we could do installs with a degraded
RAID mirror ( single drive ), but I think I saw Jeremy ( could be
wrong ), explain that we could rather reserver extra superblocks, and
convert an ext3 partition to a RAID mirror, with a little less effort.

So, my question is, where could I find the information to mess with
ext3 to RAID 1 convertion and could we really look at putting this
into anaconda for installs.

Motivation - Well, we believe in having our boot and OS ( lightly
used ) drives always mirror to keep low, down time, but when we do
upgrades or re-installs, we try and keep one drive out of the install
to recover conf files, setting and the like and once the install is
running ( a day or two ), we then add the drive back into the RAID and
sync. At the moment, we have to take a spare drive with to do
installs and upgrade and then hardware get's a little out of sync with
what might have been orginally purchased, leaving us with odd sized
drives ...

Thanks
Mailed
LeeT

_______________________________________________
Anaconda-devel-list mailing list
Anaconda-devel-list@redhat.com
https://www.redhat.com/mailman/listinfo/anaconda-devel-list
 
Old 03-11-2009, 12:49 PM
Bruno Wolff III
 
Default Mirror RAID convert ...

On Wed, Mar 11, 2009 at 09:52:45 +0200,
Clinton Lee Taylor <clintonlee.taylor@gmail.com> wrote:
>
> Was thinking or hoping, that we could do installs with a degraded
> RAID mirror ( single drive ), but I think I saw Jeremy ( could be
> wrong ), explain that we could rather reserver extra superblocks, and
> convert an ext3 partition to a RAID mirror, with a little less effort.

Note that degraded and a raid 1 array that is complete but only has
one device are different cases.

Reserving extra space for the meta data is tricky, as different amounts
can be used (in different locations) depending how raid is being set up.
If you don't know how it will eventually be set up, you'll want to make
a worst cases guess.

> So, my question is, where could I find the information to mess with
> ext3 to RAID 1 convertion and could we really look at putting this
> into anaconda for installs.

You need to change the partition code and run mdadm to create the array.
As long as the meta data doesn't overwrite the data it should work.
If the intention is to support only raid 1 (the most common case where
people might want this), then I suggest creating the raid array first
with just one element.

> Motivation - Well, we believe in having our boot and OS ( lightly
> used ) drives always mirror to keep low, down time, but when we do
> upgrades or re-installs, we try and keep one drive out of the install
> to recover conf files, setting and the like and once the install is
> running ( a day or two ), we then add the drive back into the RAID and
> sync. At the moment, we have to take a spare drive with to do
> installs and upgrade and then hardware get's a little out of sync with
> what might have been orginally purchased, leaving us with odd sized
> drives ...

If you can throw in spare drives, you can use a drive at least as big
as what's in there, add it to the mirror, let it sync and remove it
from the mirror. And then you have your backup on disk.

I typically want to be able to do this when changing the disk layout.
I don't have spare disks to plug in and don't want to do a full recovery
from DVDs or tape, when using one of my mirrors is faster.

_______________________________________________
Anaconda-devel-list mailing list
Anaconda-devel-list@redhat.com
https://www.redhat.com/mailman/listinfo/anaconda-devel-list
 
Old 03-12-2009, 08:15 AM
Clinton Lee Taylor
 
Default Mirror RAID convert ...

Greetings ...

Thanks for the quick reply ...

2009/3/11 Bruno Wolff III <bruno@wolff.to>:
> On Wed, Mar 11, 2009 at 09:52:45 +0200,
> *Clinton Lee Taylor <clintonlee.taylor@gmail.com> wrote:
>>
>> *Was thinking or hoping, that we could do installs with a degraded
>> RAID mirror ( single drive ), but I think I saw Jeremy ( could be
>> wrong ), explain that we could rather reserver extra superblocks, and
>> convert an ext3 partition to a RAID mirror, with a little less effort.
>
> Note that degraded and a raid 1 array that is complete but only has
> one device are different cases.

Okay, was really only referring to degraded raid 1 MD which would
require another device ( partition ) to be complete.

> Reserving extra space for the meta data is tricky, as different amounts
> can be used (in different locations) depending how raid is being set up.
> If you don't know how it will eventually be set up, you'll want to make
> a worst cases guess.

I will admit, I have not studied the on-disk format of ext2/3/4 or
RAID1, the last time I did this, was in the good old days of DOS and
FAT12/16 ... We really looking for only one case, RAID1, I don't see a
need for anything else, in this topic's reguard ... Does RAID1 require
twice as many superblocks as a standard ext3 partition? That is if we
are talking about making a dirty conversion problem ...

Part of this comes from M$ world, where a server can be installed
with one disk and upgraded to second disk later, and bound into a
mirror ... Would be great, if we could come up with a similar system,
that is not all hackish ... Found a nice howto on booting using rescue
disk and then creating degraded arrary to move running system onto and
then rebooting into OS to added second drive into to mirror ...

This might not yet be the right time to wish away my life ( no flame
war or anything ), but if we added the extra superblocks into the
filesystem and populated with the correct data, changed the partition
ID to fd (Linux raid auto), maybe generate and updated initrd and
reboot said server with extra hard drive in ... And presto, new
degraded RAID1 setup ready to hot-add new spare md devices ... Yum
jokes aside, this would a dream world for me, but I'm hoping that we
could work towards that ...

>> *So, my question is, where could I find the information to mess with
>> ext3 to RAID 1 conversion and could we really look at putting this
>> into anaconda for installs.
>
> You need to change the partition code and run mdadm to create the array.

This I would think would be to create a GUI interface ... Thinking
maybe just messing with the KickStart system first and then grow out
to include GUI/TUI options ... mdadm is already used to create the md
devices, so if we add an extra option, to create a degraded raid1
partition, then I believe that part of my problem would be solve ...

> As long as the meta data doesn't overwrite the data it should work.
I'm not sure what you referring to know ... This comment seems out of
place, but I could just be missing understanding what you getting at
...

> If the intention is to support only raid 1 (the most common case where
> people might want this), then I suggest creating the raid array first
> with just one element.

I think you are agreeing with me, or at least as I see it with my
limit knowledge ...

>> *Motivation - Well, we believe in having our boot and OS ( lightly
>> used ) drives always mirror to keep low, down time, but when we do
>> upgrades or re-installs, we try and keep one drive out of the install
>> to recover conf files, setting and the like and once the install is
>> running ( a day or two ), we then add the drive back into the RAID and
>> sync. *At the moment, we have to take a spare drive with to do
>> installs and upgrade and then hardware get's a little out of sync with
>> what might have been originally purchased, leaving us with odd sized
>> drives ...
>
> If you can throw in spare drives, you can use a drive at least as big
> as what's in there, add it to the mirror, let it sync and remove it
> from the mirror. And then you have your backup on disk.

Well, this is how we are going it at the moment, but an upgrade the
costs us an extra drive and mirroring twice per upgrade, which is what
I'm trying to get around ... Backup of already installed system has a
very limited use at the moment, but can be used as a selling point to
the point hair people ...

> I typically want to be able to do this when changing the disk layout.
> I don't have spare disks to plug in and don't want to do a full recovery
> from DVDs or tape, when using one of my mirrors is faster.

Again, it sounds like you understand my pain ...

Maybe this is really two ideas and that we should only be disuccing
ananconda here, but I'm sure Jeremy said that he had done the reserver
the needed superblocks to do a conversion and would be a worthwhile
trade off, if this was a working idea ... If anybody thinks this is
off topic, then I will try and chat with Niel Brown, seeing that I
believe that he might be able to help, but I would rather be seeing
this in Fedora, seeing that I am Fedora user and could help, even a
little ...

Thanks
Mailed
LeeT

_______________________________________________
Anaconda-devel-list mailing list
Anaconda-devel-list@redhat.com
https://www.redhat.com/mailman/listinfo/anaconda-devel-list
 
Old 03-12-2009, 02:33 PM
Bruno Wolff III
 
Default Mirror RAID convert ...

On Thu, Mar 12, 2009 at 11:15:38 +0200,
Clinton Lee Taylor <clintonlee.taylor@gmail.com> wrote:
>
> I will admit, I have not studied the on-disk format of ext2/3/4 or
> RAID1, the last time I did this, was in the good old days of DOS and
> FAT12/16 ... We really looking for only one case, RAID1, I don't see a
> need for anything else, in this topic's reguard ... Does RAID1 require
> twice as many superblocks as a standard ext3 partition? That is if we
> are talking about making a dirty conversion problem ...

If you are just worried about raid 1, I would recommend not supporting
degraded arrays, but rather allow raid 1 arrays with only 1 element.
This will be safer and doing the prep work isn't too hard.

The raid arrays are independent of file system you put on top of it, except
for the partition grub accesses. For that one you need the raid metadata
to be at the end (which is normal) so that grub can ignore the raid aspect.

All that really needs to be done is to reduce the minimum number of elements
for raid 1 arrays to 1 instead of 2 and to use the --force option when
creating a new one element array.

I have done this for earlier versions of anaconda and its not too hard.
There are some patches showing how in a closed bug. Though with the recent
overhaul going on with how anaconda handles devices, they may not apply
too cleanly any more.

_______________________________________________
Anaconda-devel-list mailing list
Anaconda-devel-list@redhat.com
https://www.redhat.com/mailman/listinfo/anaconda-devel-list
 

Thread Tools




All times are GMT. The time now is 01:15 PM.

VBulletin, Copyright ©2000 - 2014, Jelsoft Enterprises Ltd.
Content Relevant URLs by vBSEO ©2007, Crawlability, Inc.
Copyright 2007 - 2008, www.linux-archive.org