Re: Unbearably slow performance

On Sat, Dec 20, 2008 at 11:55:30PM +0000, redshift@... wrote:
> Hello
> 
> I'm testing out ATA-over-Ethernet, and I'm getting very poor
> results. On the (only) client, I can't even get read speeds above 1
> megabyte per second:
...
> ifconfig shows 0 errors, drops, overruns, frame or carrier errors,
> on BOTH sides. Each side is connected using gigabit ethernet with
> the MTU set to 1500. On the target side I'm using vblade:
...
> The weird part is, I can't always reproduce this. When the stars are
> right and I modprobe aoe, I get "normal" read speeds, which are about
> 75% of what the actual disk is capable of.
...
> What else can I look at?

One thing to look for would be a network link that is switching
between gigabit and 10 Mbps.

--

-- 
  Ed Cashin <ecashin@...>

------------------------------------------------------------------------------
This SF.net email is sponsored by:
SourcForge Community
SourceForge wants to tell your story.
http://p.sf.net/sfu/sf-spreadtheword

patch for vblade: ability to drop privileges, compiler warnings

Hello,

although vblade's codebase is not very large, I felt that it was
desirable to have most of it run as a non-privileged user.

Here is a set of patches against vblade-19 that provide enable vblade
to drop root privileges after it has bound the raw socket. This
behavior can be enabled

* by installing vblade setuid root and calling it as a "normal" user
* or by specifying the user with the new `-u' switch on the command line.

I have tested this on an x86-64 Debian/sid box.

Cheers,
-Hilko

------------------------------------------------------------------------------
This SF.net email is sponsored by:
SourcForge Community
SourceForge wants to tell your story.
http://p.sf.net/sfu/sf-spreadtheword

_______________________________________________
Aoetools-discuss mailing list
Aoetools-discuss@...
https://lists.sourceforge.net/lists/listinfo/aoetools-discuss

aoe6-68 doesn't build against 2.6.28 ...

Hello Ed,

just to let you know, the build of aoe6-68 against vanilla 2.6.28 fails
with the following error:

24-kv.sh Error: could not identify kernel source state

I found out that device_create_drvdata was renamed to device_create in
the 2.6.28 cycle. There are other things that have changed in 2.6.28 but
i don't have access to my test box right now so i have to post more
details on the other changes later.

Just to check if we can abandon the in-house usage of aoe6 and stick to
the aoe module in Mainline, let me ask the following question:

What functionality is missing in mainline that is in version 68 and 
when will the changes and additions in functionality be merged back into
mainline?

Thanks in advance,

Marc

------------------------------------------------------------------------------
This SF.net email is sponsored by:
SourcForge Community
SourceForge wants to tell your story.
http://p.sf.net/sfu/sf-spreadtheword

Re: aoe6-68 doesn't build against 2.6.28 ...

On Mon, Jan 19, 2009 at 02:45:14PM +0100, Marc - A. Dahlhaus [ Administration | Westermann GmbH ] wrote:
> Hello Ed,
> 
> just to let you know, the build of aoe6-68 against vanilla 2.6.28 fails
> with the following error:
> 
> 24-kv.sh Error: could not identify kernel source state
> 
> I found out that device_create_drvdata was renamed to device_create in
> the 2.6.28 cycle. There are other things that have changed in 2.6.28 but
> i don't have access to my test box right now so i have to post more
> details on the other changes later.

Ah, yes.  Greg Kroah-Hartman said this change was bound to happen.
This part of the kernel is the most churn-filled that the aoe
compatibility system has to deal with.

Since I knew the change was coming, I prepared for it and should be
able to accomodate it in version 69 without too much delay.

> Just to check if we can abandon the in-house usage of aoe6 and stick to
> the aoe module in Mainline, let me ask the following question:
> 
> What functionality is missing in mainline that is in version 68 and 
> when will the changes and additions in functionality be merged back into
> mainline?

You can check the NEWS file in the aoe sources to see the changes that
appear after version 47.  The same file is online at the URL below.

Re: aoe6-68 doesn't build against 2.6.28 ...

Ed Cashin schrieb:
> On Mon, Jan 19, 2009 at 02:45:14PM +0100, Marc - A. Dahlhaus [ Administration | Westermann GmbH ] wrote:
>   
>> Hello Ed,
>>
>> just to let you know, the build of aoe6-68 against vanilla 2.6.28 fails
>> with the following error:
>>
>> 24-kv.sh Error: could not identify kernel source state
>>
>> I found out that device_create_drvdata was renamed to device_create in
>> the 2.6.28 cycle. There are other things that have changed in 2.6.28 but
>> i don't have access to my test box right now so i have to post more
>> details on the other changes later.
>>     
>
> Ah, yes.  Greg Kroah-Hartman said this change was bound to happen.
> This part of the kernel is the most churn-filled that the aoe
> compatibility system has to deal with.
>
> Since I knew the change was coming, I prepared for it and should be
> able to accomodate it in version 69 without too much delay.
>   
Some other stuff is:
'struct gendisk' lost the dev member so you have to use the disk_to_dev 
helper
'struct gendisk' lost the capacity member so you have to use the 
{get,set}_capacity helpers
Also the disk_stat things are gone now, the corresponding patch is this one:
https://kerneltrap.org/mailarchive/linux-kernel/2008/8/25/3067064

Re: aoe6-68 doesn't build against 2.6.28 ...

I have to make sure the changes work on a few popular kernels before
making a release, but I think this patch against version 68 resolves
the compatibility issues for 2.6.28.

  http://support.coraid.com/support/linux/tmp/aoe6-68-69pre1.diff

Please try it out and let me know how it works for you.

I am going to try to make time to get the kernel.org driver caught
up.  The coraid.com one has been changing a lot lately, and the
kernel.org development model has too, with the addition of linux-next
and such.  Things seem to be calming down, though.

--

-- 
  Ed Cashin <ecashin@...>

------------------------------------------------------------------------------
This SF.net email is sponsored by:
SourcForge Community
SourceForge wants to tell your story.
http://p.sf.net/sfu/sf-spreadtheword

Re: aoe6-68 doesn't build against 2.6.28 ...

On Wed, Jan 21, 2009 at 03:12:32PM -0500, Ed Cashin wrote:
> I have to make sure the changes work on a few popular kernels before
> making a release, but I think this patch against version 68 resolves
> the compatibility issues for 2.6.28.

A new prerelease fixes the conf/21 test and adds compatibility with
the new ioctl-handler prototype in 2.6.28.

  http://support.coraid.com/support/linux/tmp/aoe6-68-69pre1.diff

--

-- 
  Ed Cashin <ecashin@...>

------------------------------------------------------------------------------
This SF.net email is sponsored by:
SourcForge Community
SourceForge wants to tell your story.
http://p.sf.net/sfu/sf-spreadtheword