Which journal relocation+datalogging patches?

Which journal relocation+datalogging patches?

[JP Howard]

I'm trying to build a 2.4.20pre1 kernel with data logging patches,
journal relocation, and and important performance/reliability patches.

I see from
ftp://ftp.suse.com/pub/people/mason/patches/data-logging/2.4.20/README
that these patches are against the journal relocation patches. But where
are they? I've found a patch against 2.5.3, but that doesn't apply
cleanly to 2.4.20pre1.

If this patch doesn't yet exist against 2.4.20, could you please upload
it someplace Chris, and let me know what order to apply the patches?
(Since you obviously have made it, to create the data logging patch
against it! :-)

I looks to me like all other important patches are already in
2.4.20pre1--is that right?

What are the steps then required to move the journal to an external
device (we're using Umem NVRAM, FWIW)? How stable are the journal
relocation patches? If they're not recommended for production as yet,
would we be able to move the journal to an external device later, without
rebooting (or at least without recreating the filesystem)?

I'm really looking forward to trying this out--I plan to set aside an
identical server for testing and benchmarking, so hopefully we'll be able
to provide some interesting benchmarks of IMAP server performance with
external journals.

TIA,
  Jeremy
  http://fastmail.fm

Re: Which journal relocation+datalogging patches?

[Chris Mason]

On Wed, 2002-11-13 at 22:30, JP Howard wrote:
> I'm trying to build a 2.4.20pre1 kernel with data logging patches,
> journal relocation, and and important performance/reliability patches.
> 
> I see from
> ftp://ftp.suse.com/pub/people/mason/patches/data-logging/2.4.20/README
> that these patches are against the journal relocation patches. But where
> are they? I've found a patch against 2.5.3, but that doesn't apply
> cleanly to 2.4.20pre1.
> 
> If this patch doesn't yet exist against 2.4.20, could you please upload
> it someplace Chris, and let me know what order to apply the patches?
> (Since you obviously have made it, to create the data logging patch
> against it! :-)

Sorry for the confusion.  The journal relocation parts are in
01-relocation-5.diff.gz, the README was attempting to make it clear
namesys authored it.  Hopefully it is a little clearer now.

> 
> I looks to me like all other important patches are already in
> 2.4.20pre1--is that right?
> 
> What are the steps then required to move the journal to an external
> device (we're using Umem NVRAM, FWIW)? How stable are the journal
> relocation patches? If they're not recommended for production as yet,
> would we be able to move the journal to an external device later, without
> rebooting (or at least without recreating the filesystem)?

You can use reiserfstune to change the log device of an existing disk. 
The relocation and data logging patches are considered stable, both are
included in the suse kernels.

To try the external logging device mkreisefs -j /dev/xxx -s 8192
/dev/yyy

That will create a 32MB log on /dev/xxx, and the main filesystem on
/dev/yyy

> 
> I'm really looking forward to trying this out--I plan to set aside an
> identical server for testing and benchmarking, so hopefully we'll be able
> to provide some interesting benchmarks of IMAP server performance with
> external journals.

Excellent, I'd love to see those numbers.

-chris

Re: Which journal relocation+datalogging patches?

[JP Howard]

On 14 Nov 2002 10:10:14 -0500, "Chris Mason" <mason@suse.com> said:
> Sorry for the confusion.  The journal relocation parts are in
> 01-relocation-5.diff.gz, the README was attempting to make it clear
> namesys authored it.  Hopefully it is a little clearer now.
> 
Perfectly clear, thankyou.

> You can use reiserfstune to change the log device of an existing disk. 
> The relocation and data logging patches are considered stable, both are
> included in the suse kernels.
> 
> To try the external logging device mkreisefs -j /dev/xxx -s 8192
> /dev/yyy
> 
> That will create a 32MB log on /dev/xxx, and the main filesystem on
> /dev/yyy
> 
Thanks for this info. I'll try this on the weekend.
> > 
> > I'm really looking forward to trying this out--I plan to set aside an
> > identical server for testing and benchmarking, so hopefully we'll be able
> > to provide some interesting benchmarks of IMAP server performance with
> > external journals.
> 
> Excellent, I'd love to see those numbers.
> 
OK, I'll be sure to report back when we have something. That may yet take
some months--our IMAP stress test tool is not yet correctly simulating
our production profile, and is resulting in CPU bound rather than IO
bound tests. So we'll have to get that working properly first, and then
we'll be able to benchmark.