reiser4 disk formats (was http://www.namesys.com/snapshots/2004.08.09-internal.testing/)

reiser4 disk formats (was http://www.namesys.com/snapshots/2004.08.09-internal.testing/)

[Domenico Andreoli]

hi all,

On Mon, Aug 09, 2004 at 11:49:31PM +0400, Vitaly Fertman wrote:
> 
> and it has a converter from the previous format to the new one -- if you 
> already created a reiser4 fs with the mkfs from some previous version 
> of reiser4progs, first of all run 
>     $ debugfs.reiser4 -C device
 
[ why i can't find any -C option in the debugfs.reiser4' manpage? it
is instead present in the help of debugfs.reiser4. probably i'll add
it to the debian manpage. ]

if i launched this command using reiser4progs 1.0.0 would it make all
the required conversions to bring up to date the disk format?

the problem is that in debian currently i have two versions of
reiser4progs, 0.5.3 and 0.5.6. in the next few days i want to upload also
1.0.0 to push it into sarge (the next debian stable due to being released
in few weeks and for which the uploads will be closed on 17th August).

hence i must explain to the user what she is required to do in order
to update her filesystems and kernel.

IIRC 0.5.3, 0.5.6 and 1.0.0 have all different disk format, haven't them?

are the following the commands required to properly convert the disk
formats?

to convert a 0.5.3 to a 0.5.6:
  $ fsck.reiser4 <device>

to convert a 0.5.6 to 1.0.0:
  $ debugfs.reiser4 -C <device>

to convert a 0.5.3 to a 1.0.0:
  $ fsck.reiser4 [--fix] <device>
  $ debugfs.reiser4 -C <device>


thanks in advance.

cheers
domenico

-----[ Domenico Andreoli, aka cavok
 --[ http://people.debian.org/~cavok/gpgkey.asc
   ---[ 3A0F 2F80 F79C 678A 8936  4FEE 0677 9033 A20E BC50

Re: reiser4 disk formats (was http://www.namesys.com/snapshots/2004.08.09-internal.testing/)

[Domenico Andreoli]

On Fri, Aug 13, 2004 at 07:18:27PM +0200, Domenico Andreoli wrote:
> On Mon, Aug 09, 2004 at 11:49:31PM +0400, Vitaly Fertman wrote:
> > 
> > and it has a converter from the previous format to the new one -- if you 
> > already created a reiser4 fs with the mkfs from some previous version 
> > of reiser4progs, first of all run 
> >     $ debugfs.reiser4 -C device
>  
> [ why i can't find any -C option in the debugfs.reiser4' manpage? it
> is instead present in the help of debugfs.reiser4. probably i'll add
> it to the debian manpage. ]
> 
> if i launched this command using reiser4progs 1.0.0 would it make all
> the required conversions to bring up to date the disk format?
> 
> the problem is that in debian currently i have two versions of
> reiser4progs, 0.5.3 and 0.5.6. in the next few days i want to upload also
> 1.0.0 to push it into sarge (the next debian stable due to being released
> in few weeks and for which the uploads will be closed on 17th August).
> 
> hence i must explain to the user what she is required to do in order
> to update her filesystems and kernel.
> 
> IIRC 0.5.3, 0.5.6 and 1.0.0 have all different disk format, haven't them?
> 
> are the following the commands required to properly convert the disk
> formats?
> 
> to convert a 0.5.3 to a 0.5.6:
>   $ fsck.reiser4 <device>
> 
> to convert a 0.5.6 to 1.0.0:
>   $ debugfs.reiser4 -C <device>
    $ fsck.reiser4 [--fix] <device>
> 
> to convert a 0.5.3 to a 1.0.0:
>   $ fsck.reiser4 [--fix] <device>
>   $ debugfs.reiser4 -C <device>
    $ fsck.reiser4 [--fix] <device>

it looks i have also been touched by the "wrong bytes" bug (wasn't
it solved??)

thanks again
domenico

-----[ Domenico Andreoli, aka cavok
 --[ http://people.debian.org/~cavok/gpgkey.asc
   ---[ 3A0F 2F80 F79C 678A 8936  4FEE 0677 9033 A20E BC50

Re: reiser4 disk formats (was http://www.namesys.com/snapshots/2004.08.09-internal.testing/)

[Vitaly Fertman]

On Friday 13 August 2004 21:18, Domenico Andreoli wrote:
> hi all,
>
> On Mon, Aug 09, 2004 at 11:49:31PM +0400, Vitaly Fertman wrote:
> > and it has a converter from the previous format to the new one -- if you
> > already created a reiser4 fs with the mkfs from some previous version
> > of reiser4progs, first of all run
> >     $ debugfs.reiser4 -C device
>
> [ why i can't find any -C option in the debugfs.reiser4' manpage? it
> is instead present in the help of debugfs.reiser4. probably i'll add
> it to the debian manpage. ]

ok. I will add it also.

> if i launched this command using reiser4progs 1.0.0 would it make all
> the required conversions to bring up to date the disk format?
>
> the problem is that in debian currently i have two versions of
> reiser4progs, 0.5.3 and 0.5.6. in the next few days i want to upload also
> 1.0.0 to push it into sarge (the next debian stable due to being released
> in few weeks and for which the uploads will be closed on 17th August).
>
> hence i must explain to the user what she is required to do in order
> to update her filesystems and kernel.
>
> IIRC 0.5.3, 0.5.6 and 1.0.0 have all different disk format, haven't them?
>
> are the following the commands required to properly convert the disk
> formats?
>
> to convert a 0.5.3 to a 0.5.6:
>   $ fsck.reiser4 <device>
>
> to convert a 0.5.6 to 1.0.0:
>   $ debugfs.reiser4 -C <device>
>
> to convert a 0.5.3 to a 1.0.0:
>   $ fsck.reiser4 [--fix] <device>
>   $ debugfs.reiser4 -C <device>

there were a few format changes in 0.5.5. The first one is the formatting policy 
id was changed, to fix this 
	fsck.reiser4 --build-sb device
is needed. The second is object plugin sets. It was supposed to run
	fsck.reiser4 --fix device
to add pset to the root directory. Actually you could merely run
	fsck.reiser4 device
and it tells you what to do next.

To jump from 0.5.3 to 1.0.0 it would be better to run:
	fsck.reiser4 --build-sb device
	debugfs.reiser4 -C device
	fsck.reiser4 --fix device
all tools are from 1.0.0 here.

The converter (debugfs.reiser4 -C) is supposed to be run on consistent fs of 
course, so probably it is worth to check/fix it with the current reiser4progs 
before any update.

-- 
Thanks,
Vitaly Fertman

Re: reiser4 disk formats (was http://www.namesys.com/snapshots/2004.08.09-internal.testing/)

[Domenico Andreoli]

On Sun, Aug 15, 2004 at 03:24:34PM +0400, Vitaly Fertman wrote:
>
> there were a few format changes in 0.5.5. The first one is the formatting policy 
> id was changed, to fix this
>   fsck.reiser4 --build-sb device
> is needed. The second is object plugin sets. It was supposed to run
>   fsck.reiser4 --fix device
> to add pset to the root directory. Actually you could merely run
>   fsck.reiser4 device
> and it tells you what to do next.
>
> To jump from 0.5.3 to 1.0.0 it would be better to run:
>   fsck.reiser4 --build-sb device
>   debugfs.reiser4 -C device
>   fsck.reiser4 --fix device
> all tools are from 1.0.0 here.
>
> The converter (debugfs.reiser4 -C) is supposed to be run on consistent fs of
> course, so probably it is worth to check/fix it with the current reiser4progs
> before any update.
>

so the most safe catch-all command sequence should be simply

  1. fsck.reiser4 device
     - do also whatever fsck says is required to bring the fs
       consistent. most probably it will say to use --build-sb and
       then --fix.
  2. debugfs.reiser4 -C device
  3. fsck.reiser4 device
     - do again whatever fsck says, most probably --fix is required.

this should bring the fs from whatever disk format it is in to the
latest one and this *should* happen without data loss. am i right?

is command 2. idempotent?

i'd like to say to my users something like this:
  
  "ok, until this moment we have been playing. if you want to play
   harder, whatever state your disk is in, the command sequence is the
   following [...]. your data should be preserved but making a backup
   copy might be a good idea."

many thanks.

cheers
domenico

-----[ Domenico Andreoli, aka cavok
 --[ http://people.debian.org/~cavok/gpgkey.asc
   ---[ 3A0F 2F80 F79C 678A 8936  4FEE 0677 9033 A20E BC50

Re: reiser4 disk formats (was http://www.namesys.com/snapshots/2004.08.09-internal.testing/)

[Francesco Biscani]

Hi,

On Monday 16 August 2004 00:55, Domenico Andreoli wrote:
> so the most safe catch-all command sequence should be simply
>
>   1. fsck.reiser4 device
>      - do also whatever fsck says is required to bring the fs
>        consistent. most probably it will say to use --build-sb and
>        then --fix.
>   2. debugfs.reiser4 -C device
>   3. fsck.reiser4 device
>      - do again whatever fsck says, most probably --fix is required.
>
> this should bring the fs from whatever disk format it is in to the
> latest one and this *should* happen without data loss. am i right?

I may be wrong here, but as far as I remember reiser4progs 1.x give errors if 
used on old disk formats. If you try to --fix these errors _before_ changing 
format you may lose data. So the best sequence would be:

1. fsck.reiser4 --fix device 
     (with reiser4progs < 1.x)
2. upgrade to reiser4progs 1.x
3. debugfs.reiser4 -C device
4. fsck.reiser4 --fix device

This way you fix all real errors (possibly) present _before_ converting the 
layout, and you avoid the (false) errors reported by reiser4progs 1.x on old 
disk layouts. Can someone confirm?

Regards,

  Francesco

Re: reiser4 disk formats (was http://www.namesys.com/snapshots/2004.08.09-internal.testing/)

[Domenico Andreoli]

On Mon, Aug 16, 2004 at 01:13:38AM +0200, Francesco Biscani wrote:
> Hi,
> 
> On Monday 16 August 2004 00:55, Domenico Andreoli wrote:
> > so the most safe catch-all command sequence should be simply
> >
> >   1. fsck.reiser4 device
> >      - do also whatever fsck says is required to bring the fs
> >        consistent. most probably it will say to use --build-sb and
> >        then --fix.
> >   2. debugfs.reiser4 -C device
> >   3. fsck.reiser4 device
> >      - do again whatever fsck says, most probably --fix is required.
> >
> > this should bring the fs from whatever disk format it is in to the
> > latest one and this *should* happen without data loss. am i right?
> 
> I may be wrong here, but as far as I remember reiser4progs 1.x give errors if 
> used on old disk formats. If you try to --fix these errors _before_ changing 
> format you may lose data. So the best sequence would be:
> 
> 1. fsck.reiser4 --fix device 
>      (with reiser4progs < 1.x)
> 2. upgrade to reiser4progs 1.x
> 3. debugfs.reiser4 -C device
> 4. fsck.reiser4 --fix device
> 
> This way you fix all real errors (possibly) present _before_ converting the 
> layout, and you avoid the (false) errors reported by reiser4progs 1.x on old 
> disk layouts. Can someone confirm?
> 

i'm talking about a message to be put in the reiser4progs 1.0.0 package.
when the user reads the message she will not have any older version
around. so this sequence dosen't suit my needs :(

if you are right about fixing things before debugfs.reiser4 may loose
data, then the right sequence is the one from Vitaly.

what is confusing me is Vitaly saying:

"The converter (debugfs.reiser4 -C) is supposed to be run on consistent
 fs of course, so probably it is worth to check/fix it with the current
 reiser4progs before any update."

probably with "... any update." he meant also the update of the
reiser4progs package to 1.0.0, doesn't him?

cheers
domenico

-----[ Domenico Andreoli, aka cavok
 --[ http://people.debian.org/~cavok/gpgkey.asc
   ---[ 3A0F 2F80 F79C 678A 8936  4FEE 0677 9033 A20E BC50

Re: reiser4 disk formats (was http://www.namesys.com/snapshots/2004.08.09-internal.testing/)

[Francesco Biscani]

Hi,

On Monday 16 August 2004 01:27, Domenico Andreoli wrote:
> what is confusing me is Vitaly saying:
>
> "The converter (debugfs.reiser4 -C) is supposed to be run on consistent
>  fs of course, so probably it is worth to check/fix it with the current
>  reiser4progs before any update."
>
> probably with "... any update." he meant also the update of the
> reiser4progs package to 1.0.0, doesn't him?
>

Yes, I think he meant that. Maybe you could just put a message to be printed 
when reiser4progs 1.x is being installed: "Make sure you checked your reiser4 
partition(s) if upgrading to this version of reiser4progs from 0.5.x: 
avoiding to do this may lead to data loss!". This could be made interactive 
(i.e., "Type 'Yes' to continue"). 

Just a thought :)

Regards,

  Francesco

Re: reiser4 disk formats (was http://www.namesys.com/snapshots/2004.08.09-internal.testing/)

[Domenico Andreoli]

On Mon, Aug 16, 2004 at 01:38:30AM +0200, Francesco Biscani wrote:
> Hi,
> 
> On Monday 16 August 2004 01:27, Domenico Andreoli wrote:
> > what is confusing me is Vitaly saying:
> >
> > "The converter (debugfs.reiser4 -C) is supposed to be run on consistent
> >  fs of course, so probably it is worth to check/fix it with the current
> >  reiser4progs before any update."
> >
> > probably with "... any update." he meant also the update of the
> > reiser4progs package to 1.0.0, doesn't him?
> >
> 
> Yes, I think he meant that. Maybe you could just put a message to be printed 
> when reiser4progs 1.x is being installed: "Make sure you checked your reiser4 
> partition(s) if upgrading to this version of reiser4progs from 0.5.x: 
> avoiding to do this may lead to data loss!". This could be made interactive 
> (i.e., "Type 'Yes' to continue"). 
> 

i already tried this way but users did not like the installation being
interrupted. currently i have no time to make the thing polite, i have
to make the upload in the next two days otherwise it risks to miss the
next stable release.

> Just a thought :)

grazie comunque ;) 

ciao
dom

-----[ Domenico Andreoli, aka cavok
 --[ http://people.debian.org/~cavok/gpgkey.asc
   ---[ 3A0F 2F80 F79C 678A 8936  4FEE 0677 9033 A20E BC50

Re: reiser4 disk formats (was http://www.namesys.com/snapshots/2004.08.09-internal.testing/)

[Vitaly Fertman]

On Monday 16 August 2004 03:13, Francesco Biscani wrote:
> Hi,
>
> On Monday 16 August 2004 00:55, Domenico Andreoli wrote:
> > so the most safe catch-all command sequence should be simply
> >
> >   1. fsck.reiser4 device
> >      - do also whatever fsck says is required to bring the fs
> >        consistent. most probably it will say to use --build-sb and
> >        then --fix.
> >   2. debugfs.reiser4 -C device
> >   3. fsck.reiser4 device
> >      - do again whatever fsck says, most probably --fix is required.
> >
> > this should bring the fs from whatever disk format it is in to the
> > latest one and this *should* happen without data loss. am i right?
>
> I may be wrong here, but as far as I remember reiser4progs 1.x give errors
> if used on old disk formats. If you try to --fix these errors _before_
> changing format you may lose data. So the best sequence would be:

right. If you want the convertion to be done without data loss you should not 
run fsck.reiser4-1.0.0 --fix/--build-fs on NOT converted partition.

> 1. fsck.reiser4 --fix device
>      (with reiser4progs < 1.x)
> 2. upgrade to reiser4progs 1.x
> 3. debugfs.reiser4 -C device
> 4. fsck.reiser4 --fix device
>
> This way you fix all real errors (possibly) present _before_ converting the
> layout, and you avoid the (false) errors reported by reiser4progs 1.x on
> old disk layouts. Can someone confirm?

1. ensure the fs is consistent with the current reiser4progs.
Actually only severe corruptions of the reiser4 storage tree are important here,
others will be fixed later in (5).
2. update reiser4progs to 1.0.0
3. fsck.reiser4 --build-sb <device>
Needed if you move from < 0.5.5. Fixes the formatting policy id in the super 
block. Needed to be run before (4) to allow fs to be openned. 
This may tell you to run fsck with some other options (--fix/--build-fs) due 
to the format change, do not run it -- do (4) instead.
4. debugfs.reiser4 -C <device>
5. fsck.reiser4 --fix <device>
Needed if you move from < 0.5.5. Adds missed pset to the root directory.
In any case check the result is a good idea.

-- 
Thanks,
Vitaly Fertman

Re: reiser4 disk formats (was http://www.namesys.com/snapshots/2004.08.09-internal.testing/)

[Vitaly Fertman]

On Monday 16 August 2004 03:27, Domenico Andreoli wrote:
> On Mon, Aug 16, 2004 at 01:13:38AM +0200, Francesco Biscani wrote:
> > Hi,
> >
> > On Monday 16 August 2004 00:55, Domenico Andreoli wrote:
> > > so the most safe catch-all command sequence should be simply
> > >
> > >   1. fsck.reiser4 device
> > >      - do also whatever fsck says is required to bring the fs
> > >        consistent. most probably it will say to use --build-sb and
> > >        then --fix.
> > >   2. debugfs.reiser4 -C device
> > >   3. fsck.reiser4 device
> > >      - do again whatever fsck says, most probably --fix is required.
> > >
> > > this should bring the fs from whatever disk format it is in to the
> > > latest one and this *should* happen without data loss. am i right?
> >
> > I may be wrong here, but as far as I remember reiser4progs 1.x give
> > errors if used on old disk formats. If you try to --fix these errors
> > _before_ changing format you may lose data. So the best sequence would
> > be:
> >
> > 1. fsck.reiser4 --fix device
> >      (with reiser4progs < 1.x)
> > 2. upgrade to reiser4progs 1.x
> > 3. debugfs.reiser4 -C device
> > 4. fsck.reiser4 --fix device
> >
> > This way you fix all real errors (possibly) present _before_ converting
> > the layout, and you avoid the (false) errors reported by reiser4progs 1.x
> > on old disk layouts. Can someone confirm?
>
> i'm talking about a message to be put in the reiser4progs 1.0.0 package.
> when the user reads the message she will not have any older version
> around. so this sequence dosen't suit my needs :(
>
> if you are right about fixing things before debugfs.reiser4 may loose
> data, then the right sequence is the one from Vitaly.
>
> what is confusing me is Vitaly saying:
>
> "The converter (debugfs.reiser4 -C) is supposed to be run on consistent
>  fs of course, so probably it is worth to check/fix it with the current
>  reiser4progs before any update."
>
> probably with "... any update." he meant also the update of the
> reiser4progs package to 1.0.0, doesn't him?

right, the current progs before update.

-- 
Thanks,
Vitaly Fertman