[linux-lvm] Re: device-mapper transition

[linux-lvm] Re: device-mapper transition

[Joe Thornber]

Shane,

On Thu, Oct 10, 2002 at 03:07:58PM -0400, Shane Shrybman wrote:
> Is it fully backwards compatible? Can a move from 2.4 to 2.5 and back
> again without problems? I think it is but just wanted to make sure.

Yes

> In terms of making the transition it looks like I have to be running a
> kernel with dev-mapper in it before I can compile the user space LVM2
> tools.

This isn't quite true.  The lvm2 tools link to libdevmapper, which is
a tiny library that encapsulates the ioctl interface to the kernel.
In order to build libdevmapper (found in the device-mapper tarball)
you will need a kernel *source* tree that has been patched with
device-mapper.  You can use the --with-kernel_dir option to configure
to specify kernel source that is not in /usr/src/linux.

Don't hesitate to come and get help on irc.openprojects.net #lvm if
you need more help.

> I understand from your posts that LVM2 is basically stable but does not
> support snapshots at this time. Does it support pvmove, extend and
> shrink or are these dangerous/unsupported?

2.5 doesn't yet support snapshots, 2.4 does.  The only other major
feature missing is the pvmove tool (missing from both kernels).
Resizing is fine, if the tool is there we believe it works, don't be
afraid of running something because it might not be finished.

If you do try 2.5 feedback would be appreciated.

- Joe

Re: [linux-lvm] Re: device-mapper transition

[Steven Stillaway]

On Fri, Oct 11, 2002 at 08:52:21AM +0100, Joe Thornber wrote:
 
> If you do try 2.5 feedback would be appreciated.
> 
> - Joe
> 

After much time mostly spend trying to figure out how
bitkeeper works I have managed to get DM compiled and 
working with 2.5.41

There were no problems compiling and it seems to be
working fine so far.

The only really tricky bit has been the fact that the
lvm tools (in particular vgscan) are completely
different between the two version.  Actually LVM 2 just
seems to have one tool with a lot of symlinks.

So when I switch from 2.4.19 w/ LVM 1 to 2.5.41 w/
LVM2 I have to make sure to switch from one vgscan to
the other.

The obviously solution is to upgrade my 2.4.19 to 
LVM2, but not tonight -- it is late.

If I notice anything I will let you know.  If you
have any questions or want some data from my test
box please let me know.

Off to bed now -- later

--

Steven Stillaway
steve@stillaway.net

[linux-lvm] device-mapper & proc

[Steven Stillaway]

LVM has /proc/lvm, DM does not seem to have anything
equivalent.

Is there a reason for this?

Is this likely to change?

I ask because my start scripts look for /proc/lvm --
if I could have them check for /pro/lvm or /proc/dm
and run the appropriate command that would be useful.

Also I imagine a /proc/dm could be usefull for other
things.


-- 
Steven Stillaway
steve@stillaway.net

Re: [linux-lvm] Re: device-mapper transition

[Joe Thornber]

On Fri, Oct 11, 2002 at 02:04:46AM -0700, Steven Stillaway wrote:
> The only really tricky bit has been the fact that the
> lvm tools (in particular vgscan) are completely
> different between the two version.  Actually LVM 2 just
> seems to have one tool with a lot of symlinks.

You can use the single tool directly without creating the symlinks.

> So when I switch from 2.4.19 w/ LVM 1 to 2.5.41 w/
> LVM2 I have to make sure to switch from one vgscan to
> the other.

When I'm testing new tools I just run the new build without installing it.

eg,

cd <my_build_dir>/tools
./lvm vgchange -ay

- Joe

Re: [linux-lvm] device-mapper & proc

[Joe Thornber]

On Fri, Oct 11, 2002 at 02:24:55AM -0700, Steven Stillaway wrote:
> LVM has /proc/lvm, DM does not seem to have anything
> equivalent.
> 
> Is there a reason for this?

Yes, device-mapper is a much lower level driver, it is not specific to
LVM (eg, it knows nothing about volume groups).  The information
simply isn't present in the kernel to be published in a /proc file.

> Is this likely to change?

No.

> I ask because my start scripts look for /proc/lvm --
> if I could have them check for /pro/lvm or /proc/dm
> and run the appropriate command that would be useful.

I'm not sure what you mean by 'appropriate command', what information
are you getting from /proc/lvm that you can't get from the userland
tools ?  If you just want to see if device-mapper is present:

grep device-mapper /proc/devices

- Joe

Re: [linux-lvm] Re: device-mapper transition

[Andres Salomon]

We should probably have lvm2 tools check the driver (and libdevmapper)
version after processing arguments.  So, after making sure the user
doesn't just want to run vgchange --help, vgchange should check to make
sure that
a) there _is_ a version of dm in the kernel
b) the version in the kernel is compatible w/ it

This would lead to some better error messages, and hopefully less
questions.  Instead of:
[dilinger@wax tools]$ sudo vgchange -a y
  No volume groups found

Perhaps:
[dilinger@wax tools]$ sudo vgchange -a y
  Error: device-mapper driver not found; support must be compiled into
  the kernel!
[dilinger@wax tools]$ sudo vgchange -a y
  Error: the version of device-mapper in the kernel is incompatible with
  this lvm package; please upgrade device-mapper.

Support is already in the kernel and libdevmapper for driver version
querying; should be just an issue of adding the check to the lvm2 tools.


On Fri, Oct 11, 2002 at 02:04:46AM -0700, Steven Stillaway wrote:
[...]
> 
> The only really tricky bit has been the fact that the
> lvm tools (in particular vgscan) are completely
> different between the two version.  Actually LVM 2 just
> seems to have one tool with a lot of symlinks.
> 
> So when I switch from 2.4.19 w/ LVM 1 to 2.5.41 w/
> LVM2 I have to make sure to switch from one vgscan to
> the other.
> 
> The obviously solution is to upgrade my 2.4.19 to 
> LVM2, but not tonight -- it is late.
> 
> If I notice anything I will let you know.  If you
> have any questions or want some data from my test
> box please let me know.
> 
> Off to bed now -- later
> 
> --
> 
> Steven Stillaway
> steve@stillaway.net
> 
> 
> _______________________________________________
> linux-lvm mailing list
> linux-lvm@sistina.com
> http://lists.sistina.com/mailman/listinfo/linux-lvm
> read the LVM HOW-TO at http://tldp.org/HOWTO/LVM-HOWTO/

-- 
It's not denial.  I'm just selective about the reality I accept.
	-- Bill Watterson

Re: [linux-lvm] Re: device-mapper transition

[Alasdair G Kergon]

On Fri, Oct 11, 2002 at 06:08:39PM -0400, Andres Salomon wrote:
> Support is already in the kernel and libdevmapper for driver version
> querying; should be just an issue of adding the check to the lvm2 tools.

The checks are already there - it's just *all* the error messages that need 
reviewing:-)  I'll deal with these particular ones next week.

I'd like us to move towards the position where, like many other Unix 
command line utilities, if no messages are displayed it generally
means you can assume the command ran successfully.

Alasdair
-- 
agk@uk.sistina.com

Re: [linux-lvm] Re: device-mapper transition

[Steven Stillaway]

Hello Joe,

I just noticed on your website (http://people.sistina.com/~thornber)
that you have the 2.5 DM marked as untested for Intel SMP.

I am running it on an Intel SMP box (Abit BP6 with Dual Celeron 433)
and it seems fine.


-- 
Steven Stillaway
steve@stillaway.ca
steve@sparcbox.net

Re: [linux-lvm] Re: device-mapper transition

[Heinz J . Mauelshagen]

On Sat, Oct 12, 2002 at 12:07:03AM +0100, Alasdair G Kergon wrote:
> On Fri, Oct 11, 2002 at 06:08:39PM -0400, Andres Salomon wrote:
> > Support is already in the kernel and libdevmapper for driver version
> > querying; should be just an issue of adding the check to the lvm2 tools.
> 
> The checks are already there - it's just *all* the error messages that need 
> reviewing:-)  I'll deal with these particular ones next week.
> 
> I'd like us to move towards the position where, like many other Unix 
> command line utilities, if no messages are displayed it generally
> means you can assume the command ran successfully.

That's fine as a default which can be configured to fall back to the LVM1
given standard for the success case in order to stay compatible with all
exec output parsers.

> 
> Alasdair
> -- 
> agk@uk.sistina.com
> 
> _______________________________________________
> linux-lvm mailing list
> linux-lvm@sistina.com
> http://lists.sistina.com/mailman/listinfo/linux-lvm
> read the LVM HOW-TO at http://tldp.org/HOWTO/LVM-HOWTO/

-- 

Regards,
Heinz    -- The LVM Guy --

*** Software bugs are stupid.
    Nevertheless it needs not so stupid people to solve them ***

=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-

Heinz Mauelshagen                                 Sistina Software Inc.
Senior Consultant/Developer                       Am Sonnenhang 11
                                                  56242 Marienrachdorf
                                                  Germany
Mauelshagen@Sistina.com                           +49 2626 141200
                                                       FAX 924446
=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-