Ocfs2 users - Jul 2009 - upgrading tools from 1.4.1 to 1.4.2 @debian

Hi,



I want to upgrade ocfs2-tools in debian from 1.4.1 to 1.4.2. I managed to
compile .deb from the ocfs2-tools source (by copying the debian package
'debian' folder from the 1.4.1 debian source to the oracle source of
the, and adding an empty entry to the debian/changelog with the new version
number)



I see in 



http://oss.oracle.com/projects/ocfs2/dist/documentation/v1.2/ocfs2_faq.html#UPGRADE?



that I should upgrade the module too, but I can't find the module package in
debian, or the source for the module in oracle.



I found this:



http://oss.oracle.com/projects/ocfs2/files/source/v1.4/



but it didn't compile, and I am not sure if it is the module source there.
Errors:





make[2]: Entering directory `/usr/src/linux-headers-2.6.26-2-amd64'
? CC [M]  /tmp/ocfs2-1.4.2/fs/ocfs2/compat_should_remove_suid.o
? CC [M]  /tmp/ocfs2-1.4.2/fs/ocfs2/compat_generic_segment_checks.o
? CC [M]  /tmp/ocfs2-1.4.2/fs/ocfs2/alloc.o
? CC [M]  /tmp/ocfs2-1.4.2/fs/ocfs2/aops.o
? CC [M]  /tmp/ocfs2-1.4.2/fs/ocfs2/buffer_head_io.o
? CC [M]  /tmp/ocfs2-1.4.2/fs/ocfs2/dcache.o
? CC [M]  /tmp/ocfs2-1.4.2/fs/ocfs2/dir.o
/tmp/ocfs2-1.4.2/fs/ocfs2/dir.c: In function 'ocfs2_readdir':
/tmp/ocfs2-1.4.2/fs/ocfs2/dir.c:865: warning: passing argument 2 of
'ocfs2_dir_foreach_blk' from incompatible pointer type
? CC [M]  /tmp/ocfs2-1.4.2/fs/ocfs2/dlmglue.o
? CC [M]  /tmp/ocfs2-1.4.2/fs/ocfs2/export.o
/tmp/ocfs2-1.4.2/fs/ocfs2/export.c: In function 'ocfs2_decode_fh':
/tmp/ocfs2-1.4.2/fs/ocfs2/export.c:233: error: 'struct
export_operations' has no member named 'find_exported_dentr
y'
/tmp/ocfs2-1.4.2/fs/ocfs2/export.c: At top level:
/tmp/ocfs2-1.4.2/fs/ocfs2/export.c:242: error: unknown field 'decode_fh'
specified in initializer
/tmp/ocfs2-1.4.2/fs/ocfs2/export.c:242: warning: initialization from
incompatible pointer type
/tmp/ocfs2-1.4.2/fs/ocfs2/export.c:246: error: unknown field
'get_dentry' specified in initializer
/tmp/ocfs2-1.4.2/fs/ocfs2/export.c:246: warning: excess elements in struct
initializer
/tmp/ocfs2-1.4.2/fs/ocfs2/export.c:246: warning: (near initialization for
'ocfs2_export_ops')
make[4]: *** [/tmp/ocfs2-1.4.2/fs/ocfs2/export.o] Error 1
make[3]: *** [/tmp/ocfs2-1.4.2/fs/ocfs2] Error 2
make[2]: *** [_module_/tmp/ocfs2-1.4.2/fs] Error 2
make[2]: Leaving directory `/usr/src/linux-headers-2.6.26-2-amd64'
make[1]: *** [build-modules] Error 2
make[1]: Leaving directory `/tmp/ocfs2-1.4.2/fs'
make: *** [fs] Error 2



-------------- next part --------------
An HTML attachment was scrubbed...
URL:
http://oss.oracle.com/pipermail/ocfs2-users/attachments/20090704/61cd3090/attachment.html

Hi, 









Thanks for your reply, and explanations. I did? 'dmesg | grep -i ocfs2'
and indeed it is 1.5.0 , so it is probably compatible with all the new fetures
in 1.4.2, since 1.4.2<1.5 ... unless it is really and odd dev version which
would mean ... 1.4 <->1.5 ?, 1.4.1<->1.5.1 ... as mainline <->
its-beta-dev-version ? Which version do you have in kernel 2.6.29>= ?











About the snapshots, I use lvm (not ? EqualLogic), and I can make the snapshots,
it just doesn't let me mount it because it thinks that it is the same volume
twice on the same node, and tunefs doesn't let you touch the volume while it
is in use (anywhere).In 1.4.2 there is a solution for that.





-----Original Message-----

From: Brian Kroth <bpkroth at gmail.com>

To: sylarrrrrrr at aim.com

Sent: Sat, Jul 4, 2009 10:42 pm

Subject: Re: [Ocfs2-users] upgrading tools from 1.4.1 to 1.4.2 @debian














Hmm, I meant to send that to the whole list.  I let you bounce this one
if you care.

sylarrrrrrr at aim.com <sylarrrrrrr at aim.com> 2009-07-04
14:41:
>    Hi,
>    Thanks for your reply, and your bug, which had one link that was new to
>    me.

>    About the kernel, I have the current lenny 2.6.26-2-amd64 ... I
>    don't really know which module it use. I thought since my ocfs2
>    version is 1.4.1 and probably next k
ernels also have only 1.4.1
>    since 1.4.2 is very new, then there will be no difference in this
>    regard between the kernels. But you seem to know more then I about
>    this, so I would like to hear more about it. 
So the versions that you're seeing here are the versions for the
"enterprise" kernels.  (I really don't like the term which is why
I put
it in quotes).  Here's my take on it, though I'm sure it's different
from the party line or what was intended.  Basically, since OCFS2 is in
the mainline now, they kind of treat that as their second stage of
testing after themselves.  Once things have matured enough there, a new
version is released to the enterprise kernels.  That is, the changes are
backported to whatever upstream kernel release RHEL or SLES is running.
Last I checked there's were based off of 2.6.18.  Basically what it all
means is that the mainline kernel tends to have the more recent fs
features than the "versioned" ones.  You can check what version of the
fs your kernel has in it by doing the following:

modprobe ocfs2
dmesg | grep -i ocfs2

Typically, it'll be 1.5.x.  I assume this is similar to how Apache (and
the kernel used to) do things with their versioning - even is considered
stable, and odd is considered the dev tree.

The major difference between the upstream 1.2 and 1.4 lines (mainline's
1.3 and 1.5) is that 1.2 and 1.4 are network protocol incompatible.
T
hat is to say, you can't have nodes with both 1.2.x and 1.4.x running
in the same cluster because they can't communicate.  But you can have
nodes running 1.4.1 and 1.4.2 running side by side.  All this means is
that you have to take them all offline to upgrade from 1.2 to 1.4.

All that said, Oracle does a very good job of testing things and only
once have I ever run into a problem with running the mainline kernel
[1].  For that reason I would suggest that you use >= 2.6.29.  I only
had this problem on one cluster doing mail (lots of little IO).  The
other couple clusters I've been running for a couple years now do web,
video streaming, encoding, file serving, etc.  I'm a big fan of the fs
especially with the upcoming and recent developments.

Some further notes about versions.  All that above was about kernel and
fs versions which are separate from the tools versions.  It has been the
case in the past that the tools version seemed to lag a bit behind the
fs version.  That is, the mainline fs has supported features that
couldn't be used yet because there wasn't support in the tools to mount
it with that option.  Irritating, but that's the way it goes.  That is
not to say that you must have the latest kernel to use the more recent
tools.  You just won't be able to make use of the extra features that it
now supports.
>    I also have some modules that are compiled for my kernel, which
> =2
0  means that upgrading the kernel is a bit of a trouble (drbd
module,
>    which I have upgraded to 8.3.1 and forgot by now where I got the
>    module source for that... maybe from testing).
I'm guessing something like this will help you out, whatever kernel you
happen to be running:

aptitude install drbd-source module-assistant
module-assistant -t auto-install drbd-source
> I also got a reply from the developers list and he said that : "Tools
> is enough. It supports all existing versions of the kernel fs. " 0A
> So, my kernel will probably be enough, but still I wonder what I am
> missing on, my main concern is reliability, and the reason I upgrade
> is for the new feature of tunefs that lets you mount snapshots for
> backup purposes.
As I mentioned before, I think the stability in the mainline is just
fine.  However, depending upon your needs you may need to look into an
"enterprise" kernel where you will give given the support you pay for
with it.

With respect to you comment about snapshots, you don't need tools
support for that.  If you're using an EqualLogic array, I've written a
script to do this.  It would be fairly trivial to make it work with
another type of array.  I also intended to add LVM support to it and
start hosting it on a publicly available site, but just haven't gotten
around to it yet.

Brian

[1] http://oss.oracle.com/pipermail/ocfs2-users/2009-January/003218.html

>    -----Original Message-----
>    From: Brian Kroth <bpkroth at gmail.com>
>    To: sylarrrrrrr at aim.com
>    Sent: Sat, Jul 4, 2009 5:39 pm
>    Subject: Re: [Ocfs2-users] upgrading tools from 1.4.1 to 1.4.2 @debian
> 
>  Heh, I was thinking about rolling my own deb soon as well.
>  http://bugs.debian.org/cgi-bin/bugreport.cgi?bug=535471
> 
>  I pretty sure that note about updating the modules is only for users
>  with "enterprise" kernels (eg: RHEL, SLES).  For Debian and
others the
>  modules are available as part of the mainline kernel, so in your case
>  it's simply a matter of upgrading the kernel if you want to make use
of
>  the more recent fs features.  lenny-backports has 2.6.29 though I've
>  been more
>  interested to try 2.6.30 as it added a number of nice
>  features like improved allocation layout and indexed dirs.  Still
>  waiting for indexed dir support in the tools though.
> 
>  Brian
> 
>  sylarrrrrrr at aim.com <sylarrrrrrr at aim.com> 2009-07-04 05:47:
>  >    Hi,
>  >    I want to upgrade ocfs2-tools in debian from 1.4.1 to 1.4.2. I
managed
to
>  >    compile .deb from the ocfs2-tools source (by copying the debian
package
>  >    'debian' folder from the 1.4.1 debian source to the oracle
source of
the,
>  >    and adding an empty entry to the debian/changelog with the new
version
>  >    number)
>  >    I see in
>  >  =2
0
http://oss.oracle.com/projects/ocfs2/dist/documentation/v1.2/ocfs2_faq.html#UPGRADE?
>  >    that I should upgrade the module too, but I can't find the
module
package
>  >    in debian, or the source for the module in oracle.
>  >    I found this:
>  >    http://oss.oracle.com/projects/ocfs2/files/source/v1.4/
>  >    but it didn't compile, and I am not sure if it is the module
source
there.



 



-------------- next part --------------
An HTML attachment was scrubbed...
URL:
http://oss.oracle.com/pipermail/ocfs2-users/attachments/20090705/1a3e9ef7/attachment-0001.html