Re: Software RAID support

2004-05-10 Thread Anton Zinoviev
On 10.V.2004 at 17:57 (+0100) Martin Michlmayr wrote:
> 
> Hmm, maybe the libparted patch for LVM2 devices should be reverted

No, it fixes a bug in libparted.

> and they be created by hand? 

If there is some need to do this it is possible.  The script
init.d/parted from the package partman can be changed to accept only
devices having the form /dev/ide/* or /dev/scsi/*.  Actually in the
very first version of partman (it was not in the CVS) this script
filtered the other devices and accepted only IDE and SCSI.  I don't
remember why later I decided to allow other devices but for LVM this
works (for now).

> Or libparted fixed to show MD devices, I guess.

Even if we don't need to do this, it would fix in my opinion a bug in
libparted.

Anton Zinoviev


-- 
To UNSUBSCRIBE, email to [EMAIL PROTECTED]
with a subject of "unsubscribe". Trouble? Contact [EMAIL PROTECTED]



Re: Software RAID support

2004-05-10 Thread Martin Michlmayr
* Anton Zinoviev <[EMAIL PROTECTED]> [2004-05-10 19:54]:
> MD devices.  You have two choices: 1. see why parted_devices doesn't
> work for MD (the bug is somewhere either in libparted or in the
> kernel)

It's a bug in libparted.  While libparted finds (most) LVM1 devices,
it doesn't find LVM2 devices by default; however, this has been fixed
by a patch from Andres Salomon (see #247174).

> I intended to create the device directories for LVM in the same way
> you do in partman-md but libparted did this part of the work for me.

Hmm, maybe the libparted patch for LVM2 devices should be reverted and
they be created by hand?  Or libparted fixed to show MD devices, I
guess.
-- 
Martin Michlmayr
[EMAIL PROTECTED]


-- 
To UNSUBSCRIBE, email to [EMAIL PROTECTED]
with a subject of "unsubscribe". Trouble? Contact [EMAIL PROTECTED]



Re: Software RAID support

2004-05-10 Thread Anton Zinoviev
On 10.V.2004 at 17:46 (+0200) Paul Fleischer wrote:
> 
> Anton, could you explain how the lvm devices are added in the /v/l/p/devices 
> directory? It looks like being part of the core partman system, but I can't 
> really figure out what it is. Whatever it is, shouldn't it be picking up 
> md-devices too?

The script init.d/parted from package parted creates them.  The only
difference between LVM and MD is that the short program parted_devices
(which Marthin has quoted) lists the LVM devices but doesn't list the
MD devices.  You have two choices: 1. see why parted_devices doesn't
work for MD (the bug is somewhere either in libparted or in the
kernel) or 2. use a script mk-devices as you do now.

I was surprised when I saw for first time in an installation report
that partman recognised the LVM devices -- this was before the package
partman-lvm was created.  I intended to create the device directories
for LVM in the same way you do in partman-md but libparted did this
part of the work for me.

Anton Zinoviev


-- 
To UNSUBSCRIBE, email to [EMAIL PROTECTED]
with a subject of "unsubscribe". Trouble? Contact [EMAIL PROTECTED]



Re: Software RAID support

2004-05-10 Thread Paul Fleischer
On Saturday 08 May 2004 18:44, Anton Zinoviev wrote:
> On  7.V.2004 at 00:48 (+0100) Martin Michlmayr wrote:
> > Ok, /lib/partman/initd.d/35dump hangs.
> >
> > /lib/partman/init.d/35dump: IN: DUMP =dev=md=0
> > parted_server: Read command: DUMP
> > parted_server: The device =dev=md=0 is not opened.
> > parted_server: Line 1087. CRITICAL ERROR!!!  EXITING.
> >
> > So after I remove /dev/md/0, /var/lib/partman/devices/=dev=md=0 still
> > exists and 35dump doesn't like this.  Who has to remove this
> > directory?  I looked in partman-lvm for reference, but couldn't find
> > anything.  Anton?
>
> This is a bug that didn't exist in beta3...

Anton, could you explain how the lvm devices are added in the /v/l/p/devices 
directory? It looks like being part of the core partman system, but I can't 
really figure out what it is. Whatever it is, shouldn't it be picking up 
md-devices too?

Thanks,
-- 
Paul Fleischer // ProGuy

PGP key fingerprint: 755A 9FB3 F7E4 DB62 8154  C5D6 381B BBCD 7BE1 FF30


-- 
To UNSUBSCRIBE, email to [EMAIL PROTECTED]
with a subject of "unsubscribe". Trouble? Contact [EMAIL PROTECTED]



Re: Software RAID support

2004-05-08 Thread Anton Zinoviev
On  7.V.2004 at 00:48 (+0100) Martin Michlmayr wrote:
> 
> Ok, /lib/partman/initd.d/35dump hangs.
> 
> /lib/partman/init.d/35dump: IN: DUMP =dev=md=0
> parted_server: Read command: DUMP
> parted_server: The device =dev=md=0 is not opened.
> parted_server: Line 1087. CRITICAL ERROR!!!  EXITING.
> 
> So after I remove /dev/md/0, /var/lib/partman/devices/=dev=md=0 still
> exists and 35dump doesn't like this.  Who has to remove this
> directory?  I looked in partman-lvm for reference, but couldn't find
> anything.  Anton?

This is a bug that didn't exist in beta3...

I think I will be able to commit a fix after no more than a couple of hours.

Anton Zinoviev


-- 
To UNSUBSCRIBE, email to [EMAIL PROTECTED]
with a subject of "unsubscribe". Trouble? Contact [EMAIL PROTECTED]



Re: Software RAID support

2004-05-08 Thread Paul Fleischer
On Friday 07 May 2004 01:48, Martin Michlmayr wrote:
> Ok, /lib/partman/initd.d/35dump hangs.
>
> /lib/partman/init.d/35dump: IN: DUMP =dev=md=0
> parted_server: Read command: DUMP
> parted_server: The device =dev=md=0 is not opened.
> parted_server: Line 1087. CRITICAL ERROR!!!  EXITING.

Ahh, yeah, I remember seeing this.

> So after I remove /dev/md/0, /var/lib/partman/devices/=dev=md=0 still
> exists and 35dump doesn't like this.  Who has to remove this
> directory?  

I think partman-md should remove it, I'll take a look at it, as soon as 
possible (having some issues with the build_cdrom_el-torito target right 
now).

> I looked in partman-lvm for reference, but couldn't find
> anything.  Anton?

partman-lvm does things a different way, as far as I can see. The LVM devices 
appear to be added by some parts of the core partman system, but I still 
haven't figured out which ;-)

-- 
Paul Fleischer // ProGuy

PGP key fingerprint: 755A 9FB3 F7E4 DB62 8154  C5D6 381B BBCD 7BE1 FF30


-- 
To UNSUBSCRIBE, email to [EMAIL PROTECTED]
with a subject of "unsubscribe". Trouble? Contact [EMAIL PROTECTED]



Re: Software RAID support

2004-05-06 Thread Andrew Pollock
On Fri, May 07, 2004 at 09:26:23AM +1000, Herbert Xu wrote:
> On Fri, May 07, 2004 at 09:20:03AM +1000, Andrew Pollock wrote:
> > 
> > I thought it was behaviour by design. IIRC, the script explicitly grepped
> > out non-ok mirror devices.
> 
> This was fixed in 0.1.63:
> 
>   * Add inactive MD constituent devices as well (closes: #238514).

Ah. It's been a couple of months since I played with it. Cool.

Andrew


-- 
To UNSUBSCRIBE, email to [EMAIL PROTECTED]
with a subject of "unsubscribe". Trouble? Contact [EMAIL PROTECTED]



Re: Software RAID support

2004-05-06 Thread Martin Michlmayr
* tbm <[EMAIL PROTECTED]> [2004-05-05 23:56]:
> >and when I go back to partman, it hangs.  Something's wrong here.
> 
> It (repeatedly) hangs at 50%... unfortuantely, I don't have a virtual
...
> This is 100% repeatable.  You don't even need any special devices.
> Just use 1 device for RAID, go to the RAID menu, create a RAID1 device
> with 1 device, then remove it, exit and it will stop at 50% when
> starting partman.

Ok, /lib/partman/initd.d/35dump hangs.

/lib/partman/init.d/35dump: IN: DUMP =dev=md=0
parted_server: Read command: DUMP
parted_server: The device =dev=md=0 is not opened.
parted_server: Line 1087. CRITICAL ERROR!!!  EXITING.

So after I remove /dev/md/0, /var/lib/partman/devices/=dev=md=0 still
exists and 35dump doesn't like this.  Who has to remove this
directory?  I looked in partman-lvm for reference, but couldn't find
anything.  Anton?

-- 
Martin Michlmayr
[EMAIL PROTECTED]


-- 
To UNSUBSCRIBE, email to [EMAIL PROTECTED]
with a subject of "unsubscribe". Trouble? Contact [EMAIL PROTECTED]



Re: Software RAID support

2004-05-06 Thread Herbert Xu
On Fri, May 07, 2004 at 09:20:03AM +1000, Andrew Pollock wrote:
> 
> I thought it was behaviour by design. IIRC, the script explicitly grepped
> out non-ok mirror devices.

This was fixed in 0.1.63:

  * Add inactive MD constituent devices as well (closes: #238514).
-- 
Debian GNU/Linux 3.0 is out! ( http://www.debian.org/ )
Email:  Herbert Xu ~{PmV>HI~} <[EMAIL PROTECTED]>
Home Page: http://gondor.apana.org.au/~herbert/
PGP Key: http://gondor.apana.org.au/~herbert/pubkey.txt


-- 
To UNSUBSCRIBE, email to [EMAIL PROTECTED]
with a subject of "unsubscribe". Trouble? Contact [EMAIL PROTECTED]



Re: Software RAID support

2004-05-06 Thread Andrew Pollock
On Fri, May 07, 2004 at 07:59:19AM +1000, Herbert Xu wrote:
> Andrew Pollock <[EMAIL PROTECTED]> wrote:
> > 
> >> 2) The initrd image generated by mkinitrd only activates the first md-device 
> >> for some weird reason.
> > 
> > mkinitrd is fussy, the mirror must be fully built and healthy to be used.
> 
> It shouldn't be.  It's meant to pick up whatever disks are listed in the
> superblocks or raidtab.
> 
> If that's not the case, please file a bug.

What I'm saying is I've found from experience with RAID1 (using mdadm) that
if the MD device the root filesystem was on was not in a healthy state when
mkinitrd was run, it wouldn't put the commands in the initrd to attach the
other half of the mirror on bootup. So if the mirror was still syncronising
at mkinitrd run time, or the mirror was only built with one member, that's
the way the mirror would be on subsequent reboots.

I thought it was behaviour by design. IIRC, the script explicitly grepped
out non-ok mirror devices.

Andrew


-- 
To UNSUBSCRIBE, email to [EMAIL PROTECTED]
with a subject of "unsubscribe". Trouble? Contact [EMAIL PROTECTED]



Re: Software RAID support

2004-05-06 Thread Herbert Xu
Andrew Pollock <[EMAIL PROTECTED]> wrote:
> 
>> 2) The initrd image generated by mkinitrd only activates the first md-device 
>> for some weird reason.
> 
> mkinitrd is fussy, the mirror must be fully built and healthy to be used.

It shouldn't be.  It's meant to pick up whatever disks are listed in the
superblocks or raidtab.

If that's not the case, please file a bug.
-- 
Debian GNU/Linux 3.0 is out! ( http://www.debian.org/ )
Email:  Herbert Xu ~{PmV>HI~} <[EMAIL PROTECTED]>
Home Page: http://gondor.apana.org.au/~herbert/
PGP Key: http://gondor.apana.org.au/~herbert/pubkey.txt


-- 
To UNSUBSCRIBE, email to [EMAIL PROTECTED]
with a subject of "unsubscribe". Trouble? Contact [EMAIL PROTECTED]



Re: Software RAID support

2004-05-06 Thread Andrew Pollock
On Tue, May 04, 2004 at 07:35:38PM +0200, Paul Fleischer wrote:
> On Sunday 02 May 2004 03:16, Martin Michlmayr wrote:
> > Paul,
> >
> > Now that beta4 is out, what are your plans of getting mdcfg and
> > partman-md uploaded to get some testing?
> 
> As I am not sure how to go about it, I have no plans. But I am very 
> interrested in getting them uploaded, so that we can get a larger testing 
> base. 
> 
> > What's the status of these programs?
> 
> Actually, they are quite useable. mdcfg currently only supports RAID1, but it 
> can easily be extended.
> 
> There are three major showstoppers for a sucessfull installation with software 
> RAID right now:
> 
> 1) Neither the grub nor the lilo installer can handle md-devices it at all. 
> They fail miserably when both /boot and / partitions are md-devices. 

You sure? When I last played with RAID1 and mdadm, I had GRUB doing funky
stuff with everything being on MD devices.
 
> 2) The initrd image generated by mkinitrd only activates the first md-device 
> for some weird reason.

mkinitrd is fussy, the mirror must be fully built and healthy to be used.

Andrew


-- 
To UNSUBSCRIBE, email to [EMAIL PROTECTED]
with a subject of "unsubscribe". Trouble? Contact [EMAIL PROTECTED]



Re: software raid support on install

2001-02-01 Thread Adam Di Carlo

David Whedon <[EMAIL PROTECTED]> writes:

> Tue, Jan 30, 2001 at 03:04:48PM -0600 wrote:
> > Has anyone been working on support for root software raid for the debian
> > installer yet?

> If your referring to debian-installer, the replacement for boot-floppies, I
> don't think anything has been done in that area.  If you are referring to
> boot-floppies, I don't know.

Yes, I'd like to support this in woody, and that means supporting this
in woody boot-floppies (since debian-installer won't be ready).

I'm not all that familiar with raid however, so I need hackers to sign
up.

> > I am interested in starting up a project to do so, and was wondering if
> > there was any existing work.
> 
> Great. Right now debian-installer can generate boot disks
[...]

Yeah, but it won't help anyone to work on that for the next release!
Work instead on boot-floppies!

-- 
.Adam Di [EMAIL PROTECTED]http://www.onShore.com/>


-- 
To UNSUBSCRIBE, email to [EMAIL PROTECTED]
with a subject of "unsubscribe". Trouble? Contact [EMAIL PROTECTED]




Re: software raid support on install

2001-01-30 Thread David Whedon

Tue, Jan 30, 2001 at 03:04:48PM -0600 wrote:
> Has anyone been working on support for root software raid for the debian
> installer yet?
If your referring to debian-installer, the replacement for boot-floppies, I
don't think anything has been done in that area.  If you are referring to
boot-floppies, I don't know.

> 
> I am interested in starting up a project to do so, and was wondering if
> there was any existing work.
> 

Great. Right now debian-installer can generate boot disks and interact in a
limited way with the user, there is much to do.  The only thing somewhat related
to software raid is the partkit module in debian-installer cvs.  It is an
incomplete module. I envision it being able to handle all our partitioning
needs, I'm guessing that software raid would fall into that, somewhere?



-David


-- 
To UNSUBSCRIBE, email to [EMAIL PROTECTED]
with a subject of "unsubscribe". Trouble? Contact [EMAIL PROTECTED]