[OpenAFS] Which file system is the best for AFS data partitions?

2007-07-13 Thread Frank Burkhardt
Hi,

On Mon, Jun 25, 2007 at 04:46:29PM -0400, Steven Jenkins wrote:

> * What is the underlying filesystem?  what features do you have enabled?  (
> e.g., the output of dumpe2fs -h or equivalent on your system)

Ok ... I replaced my beloved XFS by reiserfs (3), created a volume
containing 19 files. Removing its backup clone took 54s which is more
than 500 times faster (considered, the time needed by the operation depends
on the # of files only) than on XFS.

I'll take the chance to ask everyone about their filesystem preferences for
(namei-) AFS data partitions. I'm especially interested in things like "I
used XYfs but moved to YZfs because of XX". Please write about non-linux
servers filesystem preferences, too.

Thank you in advance,

Frank
___
OpenAFS-info mailing list
OpenAFS-info@openafs.org
https://lists.openafs.org/mailman/listinfo/openafs-info


Re: [OpenAFS] Which file system is the best for AFS data partitions?

2007-07-13 Thread Russ Allbery
Frank Burkhardt <[EMAIL PROTECTED]> writes:

> I'll take the chance to ask everyone about their filesystem preferences
> for (namei-) AFS data partitions. I'm especially interested in things
> like "I used XYfs but moved to YZfs because of XX". Please write about
> non-linux servers filesystem preferences, too.

We use ext3 because it's mainline, supported, and I simply don't trust the
other file systems to have had sufficient real-world testing and
sufficient attention paid to recovery tools.

I care more about file system consistency and reasonable recovery from
hardware and software failure than I do about the last iota of speed.

-- 
Russ Allbery ([EMAIL PROTECTED]) 
___
OpenAFS-info mailing list
OpenAFS-info@openafs.org
https://lists.openafs.org/mailman/listinfo/openafs-info


Re: [OpenAFS] Which file system is the best for AFS data partitions?

2007-07-13 Thread Robert Banz


On Jul 13, 2007, at 16:58, Russ Allbery wrote:


Frank Burkhardt <[EMAIL PROTECTED]> writes:

I'll take the chance to ask everyone about their filesystem  
preferences

for (namei-) AFS data partitions. I'm especially interested in things
like "I used XYfs but moved to YZfs because of XX". Please write  
about

non-linux servers filesystem preferences, too.


We use ext3 because it's mainline, supported, and I simply don't  
trust the

other file systems to have had sufficient real-world testing and
sufficient attention paid to recovery tools.

I care more about file system consistency and reasonable recovery from
hardware and software failure than I do about the last iota of speed.


We used to use XFS on linux as well -- though with the performance  
differences you have noticed, I'd be interested to see the benchmarks  
on XFS with/without an fsync'ing volserver & fileserver.  Those can  
be pretty fsync() intensive operations, and that could be where XFS  
is falling down.  We had a couple fileservers that we were running  
ext3 on for awhile as well, never had any problems with them to  
complain about.


Right now we're a Solaris/ZFS shop, which isn't without its  
problems.  However, its been amazingly stable/resilient/easy to  
manage -- which is where I think Linux + whateverfilesystemyoumention  
falls down.  Sometimes that can be just as important as raw performance.


-rob
___
OpenAFS-info mailing list
OpenAFS-info@openafs.org
https://lists.openafs.org/mailman/listinfo/openafs-info


[OpenAFS] OpenAFS 1.4.4 + Kernel 2.6.22.1 compile error

2007-07-13 Thread Dirk Heinrichs
Hi,

while compiling the OpenAFS kernel module for the latest Linux kernel, I get 
the following error (on Gentoo):

  CC 
[M]  
/gentoo/build/net-fs/openafs-kernel-1.4.4/work/openafs-1.4.4/src/libafs/MODLOAD-2.6.22.1-SP/rx_kmutex.o
/gentoo/build/net-fs/openafs-kernel-1.4.4/work/openafs-1.4.4/src/libafs/MODLOAD-2.6.22.1-SP/rx_kmutex.c:
 
In function 'afs_mutex_enter':
/gentoo/build/net-fs/openafs-kernel-1.4.4/work/openafs-1.4.4/src/libafs/MODLOAD-2.6.22.1-SP/rx_kmutex.c:48:
 
warning: format '%x' expects type 'unsigned int', but argument 2 has 
type 'struct afs_kmutex_t *'
/gentoo/build/net-fs/openafs-kernel-1.4.4/work/openafs-1.4.4/src/libafs/MODLOAD-2.6.22.1-SP/rx_kmutex.c:
 
In function 'afs_mutex_exit':
/gentoo/build/net-fs/openafs-kernel-1.4.4/work/openafs-1.4.4/src/libafs/MODLOAD-2.6.22.1-SP/rx_kmutex.c:69:
 
warning: format '%x' expects type 'unsigned int', but argument 2 has 
type 'struct afs_kmutex_t *'
/gentoo/build/net-fs/openafs-kernel-1.4.4/work/openafs-1.4.4/src/libafs/MODLOAD-2.6.22.1-SP/rx_kmutex.c:
 
In function 'afs_cv_wait':
/gentoo/build/net-fs/openafs-kernel-1.4.4/work/openafs-1.4.4/src/libafs/MODLOAD-2.6.22.1-SP/rx_kmutex.c:125:
 
error: 'struct task_struct' has no member named 'thread_info'
make[6]: *** 
[/gentoo/build/net-fs/openafs-kernel-1.4.4/work/openafs-1.4.4/src/libafs/MODLOAD-2.6.22.1-SP/rx_kmutex.o]
 
Error 1
make[5]: *** 
[_module_/gentoo/build/net-fs/openafs-kernel-1.4.4/work/openafs-1.4.4/src/libafs/MODLOAD-2.6.22.1-SP]
 
Error 2
make[5]: Leaving directory `/usr/src/linux-2.6.22.1'
make[4]: *** [libafs.ko] Error 2
make[4]: Leaving directory 
`/gentoo/build/net-fs/openafs-kernel-1.4.4/work/openafs-1.4.4/src/libafs/MODLOAD-2.6.22.1-SP'
make[3]: *** [linux_compdirs] Error 2
make[3]: Leaving directory 
`/gentoo/build/net-fs/openafs-kernel-1.4.4/work/openafs-1.4.4/src/libafs'
make[2]: *** [libafs] Error 2
make[2]: Leaving directory 
`/gentoo/build/net-fs/openafs-kernel-1.4.4/work/openafs-1.4.4'
make[1]: *** [build] Error 2
make[1]: Leaving directory 
`/gentoo/build/net-fs/openafs-kernel-1.4.4/work/openafs-1.4.4'
make: *** [only_libafs] Error 2

Bye...

Dirk


signature.asc
Description: This is a digitally signed message part.


Re: [OpenAFS] Which file system is the best for AFS data partitions?

2007-07-13 Thread Stephen Joyce

Frank,

I use ext3 with noatime for vice partitions and try to limit servers to 
less than 2 TB per server for servers housing research data and .5 TB per 
server for servers housing home volumes. This has worked well so far.


Outside AFS, I use xfs for filesystems over 2 TB, but as I said, I 
currently try to limit each server to no more than 2 TB. When/if I have 
more than a few TBs per server, I'd probably use xfs. ext3 is exponentially 
painful as the size increases.


My $0.02.

Cheers, Stephen
--
Stephen Joyce
Systems AdministratorP A N I C
Physics & Astronomy Department Physics & Astronomy
University of North Carolina at Chapel Hill Network Infrastructure
voice: (919) 962-7214and Computing
fax: (919) 962-0480   http://www.panic.unc.edu

 Some people make the world turn and others just watch it spin.
   -- Jimmy Buffet

On Fri, 13 Jul 2007, Frank Burkhardt wrote:


Hi,

On Mon, Jun 25, 2007 at 04:46:29PM -0400, Steven Jenkins wrote:


* What is the underlying filesystem?  what features do you have enabled?  (
e.g., the output of dumpe2fs -h or equivalent on your system)


Ok ... I replaced my beloved XFS by reiserfs (3), created a volume
containing 19 files. Removing its backup clone took 54s which is more
than 500 times faster (considered, the time needed by the operation depends
on the # of files only) than on XFS.

I'll take the chance to ask everyone about their filesystem preferences for
(namei-) AFS data partitions. I'm especially interested in things like "I
used XYfs but moved to YZfs because of XX". Please write about non-linux
servers filesystem preferences, too.

Thank you in advance,

Frank
___
OpenAFS-info mailing list
OpenAFS-info@openafs.org
https://lists.openafs.org/mailman/listinfo/openafs-info


--



___
OpenAFS-info mailing list
OpenAFS-info@openafs.org
https://lists.openafs.org/mailman/listinfo/openafs-info


[OpenAFS] Production ready?

2007-07-13 Thread Lundgren, Andrew
I looking for a production ready distributed file system.
 
I learned about AFS about 10 years ago in college, but this is the first
time I have been able to try and use it.
 
I need to set up several linux servers that will provide data to both
linux and windows 2003 client platforms.  I have read though the online
information at openafs and it seems like a good solution.
 
Now for the fun questions.  Is the windows client suitable for a
production environment?  The windows machines will in turn be serving
data to multiple clients.  (A proprietary data server that only runs on
windows will be feeding the data off box, but the data will be stored on
linux machines with the main data stores.)
 
Which version should I consider using, the latest version that the
download link leads (1.4.x) to or the version that the windows download
leads to (1.5.x)?
 
I have been looking for a good printed book to read though to learn
more, but really can't find much.  Is there a good reference/intro book
out there?
 
I have been trying to compile openAFS on SUSE 10.2 with gcc version
4.1.0 and am running into compile problems.  I ran into problems with
ifdefs on line 115.

looking at the source code, it looks like there is infact a missing if.
I guessed which one to leave and commented out a block, but the errors
continued with ifdef problems on line 460.  
 
Is this a known issue that has been resolved elsewhere?
 
Thank you!
 
--
Andrew


Re: [OpenAFS] Production ready?

2007-07-13 Thread Russ Allbery
"Lundgren, Andrew" <[EMAIL PROTECTED]> writes:

> Now for the fun questions.  Is the windows client suitable for a
> production environment?

Yes, absolutely.  The Windows client is, in fact, arguably the best of all
of the clients at the moment in terms of stability, although it has some
oddities of implementation that we hope to eventually fix.

> Which version should I consider using, the latest version that the
> download link leads (1.4.x) to or the version that the windows download
> leads to (1.5.x)?

The current recommended Windows client is the 1.5 release series.  Use 1.4
for the server and for other platforms.

> I have been looking for a good printed book to read though to learn
> more, but really can't find much.  Is there a good reference/intro book
> out there?

There isn't much that's better than the old Transarc documentation
available from the OpenAFS web site.  The documentation at the moment is
definitely a weak point.

> I have been trying to compile openAFS on SUSE 10.2 with gcc version
> 4.1.0 and am running into compile problems.  I ran into problems with
> ifdefs on line 115.

> looking at the source code, it looks like there is infact a missing if.
> I guessed which one to leave and commented out a block, but the errors
> continued with ifdef problems on line 460.  

Uh, without even knowing what file you're talking about, it's hard to
tell, but note that the 1.4.4 client release on Linux only supports Linux
kernels up through (IIRC) 2.6.20.  Later kernels will require the upcoming
1.4.5 release.

-- 
Russ Allbery ([EMAIL PROTECTED]) 
___
OpenAFS-info mailing list
OpenAFS-info@openafs.org
https://lists.openafs.org/mailman/listinfo/openafs-info


Re: [OpenAFS] Production ready?

2007-07-13 Thread Jeffrey Altman
Lundgren, Andrew wrote:
> Now for the fun questions.  Is the windows client suitable for a
> production environment? 

yes

> The windows machines will in turn be serving
> data to multiple clients.  (A proprietary data server that only runs on
> windows will be feeding the data off box, but the data will be stored on
> linux machines with the main data stores.)

ok

> Which version should I consider using, the latest version that the
> download link leads (1.4.x) to or the version that the windows download
> leads to (1.5.x)?

For Windows you should follow the instructions for Windows.

  http://www.openafs.org/windows.html

That is the production ready release of the OpenAFS client for Windows
2000 through Vista.

> I have been trying to compile openAFS on SUSE 10.2 with gcc version
> 4.1.0 and am running into compile problems.  I ran into problems with
> ifdefs on line 115.
> looking at the source code, it looks like there is infact a missing if. 
> I guessed which one to leave and commented out a block, but the errors
> continued with ifdef problems on line 460. 

I suggest using a version of Linux that has pre-built binaries available
for it unless you have some special requirements.

Jeffrey Altman



smime.p7s
Description: S/MIME Cryptographic Signature


RE: [OpenAFS] Production ready?

2007-07-13 Thread Lundgren, Andrew
 
> > Now for the fun questions.  Is the windows client suitable for a 
> > production environment?
> 
> Yes, absolutely.  The Windows client is, in fact, arguably 
> the best of all of the clients at the moment in terms of 
> stability, although it has some oddities of implementation 
> that we hope to eventually fix.
> 

Great!

> > Which version should I consider using, the latest version that the 
> > download link leads (1.4.x) to or the version that the windows 
> > download leads to (1.5.x)?
> 
> The current recommended Windows client is the 1.5 release 
> series.  Use 1.4 for the server and for other platforms.
> 

Thank you.  I was wondering if I could mix them.

> > I have been trying to compile openAFS on SUSE 10.2 with gcc version 
> > 4.1.0 and am running into compile problems.  I ran into 
> problems with 
> > ifdefs on line 115.
> 
> > looking at the source code, it looks like there is infact a 
> missing if.
> > I guessed which one to leave and commented out a block, but 
> the errors 
> > continued with ifdef problems on line 460.
> 
> Uh, without even knowing what file you're talking about, it's 
> hard to tell, but note that the 1.4.4 client release on Linux 
> only supports Linux kernels up through (IIRC) 2.6.20.  Later 
> kernels will require the upcoming
> 1.4.5 release.
> 

Lost that information in a cut and paste.  Opps.  I was able to get the
code to compile on suse by removing some extra ifdefs from the lines 115
and 460 in the file:

openafs-1.5.21/src/libafs/MODLOAD-2.6.16.27-0.9-default-SP/osi_vfsops.c

I will step back now and look at the 1.4.x code for the linux servers
though, so it won't mater.


Another question, from what I have been able to gather, the "kernel"
code refers to AFS kernel code not a custom linux kernel right?

Thanks!
___
OpenAFS-info mailing list
OpenAFS-info@openafs.org
https://lists.openafs.org/mailman/listinfo/openafs-info


Re: [OpenAFS] Production ready?

2007-07-13 Thread Russ Allbery
"Lundgren, Andrew" <[EMAIL PROTECTED]> writes:

> Lost that information in a cut and paste.  Opps.  I was able to get the
> code to compile on suse by removing some extra ifdefs from the lines 115
> and 460 in the file:

> openafs-1.5.21/src/libafs/MODLOAD-2.6.16.27-0.9-default-SP/osi_vfsops.c

> I will step back now and look at the 1.4.x code for the linux servers
> though, so it won't mater.

Oh, yeah, 1.5 is not frequently widely tested on Linux, so there may be
occasional glitches there.

> Another question, from what I have been able to gather, the "kernel"
> code refers to AFS kernel code not a custom linux kernel right?

Correct.

-- 
Russ Allbery ([EMAIL PROTECTED]) 
___
OpenAFS-info mailing list
OpenAFS-info@openafs.org
https://lists.openafs.org/mailman/listinfo/openafs-info


Re: [OpenAFS] Production ready?

2007-07-13 Thread Derrick J Brashear

On Fri, 13 Jul 2007, Lundgren, Andrew wrote:


Which version should I consider using, the latest version that the
download link leads (1.4.x) to or the version that the windows download
leads to (1.5.x)?


I'd suggest using the version the windows download link leads to on 
windows, and the latest link otherwise...



I have been trying to compile openAFS on SUSE 10.2 with gcc version
4.1.0 and am running into compile problems.  I ran into problems with
ifdefs on line 115.


line 115 of?

___
OpenAFS-info mailing list
OpenAFS-info@openafs.org
https://lists.openafs.org/mailman/listinfo/openafs-info


Re: [OpenAFS] Version 1.4.5 (was: Production ready?)

2007-07-13 Thread Dirk Heinrichs
Am Freitag, 13. Juli 2007 schrieb Russ Allbery:

> but note that the 1.4.4 client release on Linux only supports Linux
> kernels up through (IIRC) 2.6.20.

1.4.4/2.6.21.x also worked fine for me.

> Later kernels will require the upcoming 
> 1.4.5 release.

Ah, that will then solve my compilation issue with 2.6.22.1, I guess (see my 
other mail from yesterday). Will wait for 1.4.5, then.

Bye...

Dirk


signature.asc
Description: This is a digitally signed message part.