Skip to contentSkip to conversation

new reiser4 snapshot released.

ANNOUNCE Nikita Danilov 10 participants 17 messages Mar 26, 2004 – Mar 31, 2004

lore
Share & tools
lore
#0 lore raw

Hello,

new reiser4 snapshot against 2.6.5-rc2 is available at

http://www.namesys.com/snapshots/2004.03.26/

It is mainly bug-fixing release. See READ.ME for the list of fixes and caveats.

Should no significant problems be found in this snapshot, we shall start sending patches to -mm next week.

Nikita.

#1 lore raw

Reiser4 needs more testers

We have one NFS related bug remaining, and one mmap all of memory related bug (and performance issue) that you can hit using iozone. We will fix both of these in next week's snapshot, they were both multi-day bug fixes. When they are fixed, unless users/distros find bugs next week we will submit it for inclusion in the -mm and then the official kernel.

We hope it is now fairly stable for average users if you avoid those two issues (we need to get rid of those dire warnings about its stability...., we will remember that next snapshot....;-) )

We need a lot more real user testers, because we have run out of scripts that can crash it, and there are distros that would like to ship it soon. Please also complain to [email protected] and [email protected] about poor documentation, etc., ....

The new reiser4 snapshot (against 2.6.5-rc2) is available at

http://www.namesys.com/snapshots/2004.03.26/

Signature · 1 line
Hans
#2 lore raw

Re: Reiser4 needs more testers

Hans Reiser wrote:

Hi Hans,

We will test the last snapshot.

Thanks,
Giovanni

Quoted 11 lines

We have one NFS related bug remaining, and one mmap all of memory related bug (and performance issue) that you can hit using iozone. We will fix both of these in next week's snapshot, they were both multi-day bug fixes. When they are fixed, unless users/distros find bugs next week we will submit it for inclusion in the -mm and then the official kernel.

We hope it is now fairly stable for average users if you avoid those two issues (we need to get rid of those dire warnings about its stability...., we will remember that next snapshot....;-) )

We need a lot more real user testers, because we have run out of scripts that can crash it, and there are distros that would like to ship it soon. Please also complain to [email protected] and [email protected] about poor documentation, etc., ....

The new reiser4 snapshot (against 2.6.5-rc2) is available at

http://www.namesys.com/snapshots/2004.03.26/

Signature · 18 lines
-- 

--
Check FT Websites ... http://www.futuretg.com  - ftp://ftp.futuretg.com
http://www.FTLinuxCourse.com
                           /Certification
http://www.rpmparadaise.org
http://GNULinuxUtilities.com
http://www.YourPersonalOperatingSystem.com


--

      Europe:                          USA:
	Future Technologies             Future Technologies
	Viale Grigoletti, 20            1158 26th Street #592
	33170 - Pordenone (PN)          Santa Monica, CA 90403
	Italy.                          USA.
#3 lore raw

Re: Reiser4 needs more testers

In reply to Hans Reiser
Quoted 11 lines · Hans Reiser

We have one NFS related bug remaining, and one mmap all of memory related bug (and performance issue) that you can hit using iozone. We will fix both of these in next week's snapshot, they were both multi-day bug fixes. When they are fixed, unless users/distros find bugs next week we will submit it for inclusion in the -mm and then the official kernel.

We hope it is now fairly stable for average users if you avoid those two issues (we need to get rid of those dire warnings about its stability...., we will remember that next snapshot....;-) )

i assume reiser4 will be proper to use for real soon? :>

Quoted 8 lines

We need a lot more real user testers, because we have run out of scripts that can crash it, and there are distros that would like to ship it soon. Please also complain to [email protected] and [email protected] about poor documentation, etc., ....

The new reiser4 snapshot (against 2.6.5-rc2) is available at

http://www.namesys.com/snapshots/2004.03.26/

Signature · 2 lines
Regards, Redeeman
[email protected]
#4 lore raw

Re: Reiser4 needs more testers

In reply to Hans Reiser

I've run through another test on STP with the latest snapshot against reiser4 and reiserfs for comparison. Reiser4 still seems to be lagging behind reiserfs in our DBT-2 workload (the bigger number is better in the metric):

          metric  url to test results
          ------  ---------------------------------
reiser4   1208    http://khack.osdl.org/stp/290774/
reiserfs  1819    http://khack.osdl.org/stp/290775/

Mark

Quoted 19 lines · Hans Reiser

We have one NFS related bug remaining, and one mmap all of memory related bug (and performance issue) that you can hit using iozone. We will fix both of these in next week's snapshot, they were both multi-day bug fixes. When they are fixed, unless users/distros find bugs next week we will submit it for inclusion in the -mm and then the official kernel.

We hope it is now fairly stable for average users if you avoid those two issues (we need to get rid of those dire warnings about its stability...., we will remember that next snapshot....;-) )

We need a lot more real user testers, because we have run out of scripts that can crash it, and there are distros that would like to ship it soon. Please also complain to [email protected] and [email protected] about poor documentation, etc., ....

The new reiser4 snapshot (against 2.6.5-rc2) is available at

http://www.namesys.com/snapshots/2004.03.26/

#5 lore raw

Re: Reiser4 needs more testers

This was the parallel fsync benchmark?

Hans

Quoted 31 lines · [email protected] wrote:

I've run through another test on STP with the latest snapshot against reiser4 and reiserfs for comparison. Reiser4 still seems to be lagging behind reiserfs in our DBT-2 workload (the bigger number is better in the metric):

metric url to test results ------ ---------------------------------

reiser4   1208    http://khack.osdl.org/stp/290774/
reiserfs  1819    http://khack.osdl.org/stp/290775/

Mark

Quoted 11 lines · Hans Reiser

We have one NFS related bug remaining, and one mmap all of memory related bug (and performance issue) that you can hit using iozone. We will fix both of these in next week's snapshot, they were both multi-day bug fixes. When they are fixed, unless users/distros find bugs next week we will submit it for inclusion in the -mm and then the official kernel.

We hope it is now fairly stable for average users if you avoid those two issues (we need to get rid of those dire warnings about its stability...., we will remember that next snapshot....;-) )

We need a lot more real user testers, because we have run out of scripts that can crash it, and there are distros that would like to ship it soon. Please also complain to [email protected] and [email protected] about poor documentation, etc., ....

The new reiser4 snapshot (against 2.6.5-rc2) is available at

http://www.namesys.com/snapshots/2004.03.26/

Signature · 1 line
Hans
#6 lore raw

Re: Reiser4 needs more testers

It's an OLTP database workload, but it is the one with all those fsyncs that Nikita pointed out.

Mark

Quoted 49 lines · Hans Reiser

This was the parallel fsync benchmark?

Hans

Quoted 41 lines · [email protected] wrote:

I've run through another test on STP with the latest snapshot against reiser4 and reiserfs for comparison. Reiser4 still seems to be lagging behind reiserfs in our DBT-2 workload (the bigger number is better in the metric):

metric url to test results ------ ---------------------------------

reiser4   1208    http://khack.osdl.org/stp/290774/
reiserfs  1819    http://khack.osdl.org/stp/290775/

Mark

Quoted 21 lines · Hans Reiser

We have one NFS related bug remaining, and one mmap all of memory related bug (and performance issue) that you can hit using iozone. We will fix both of these in next week's snapshot, they were both multi-day bug fixes. When they are fixed, unless users/distros find bugs next week we will submit it for inclusion in the -mm and then the official kernel.

We hope it is now fairly stable for average users if you avoid those two issues (we need to get rid of those dire warnings about its stability...., we will remember that next snapshot....;-) )

We need a lot more real user testers, because we have run out of scripts that can crash it, and there are distros that would like to ship it soon. Please also complain to [email protected] and [email protected] about poor documentation, etc., ....

The new reiser4 snapshot (against 2.6.5-rc2) is available at

http://www.namesys.com/snapshots/2004.03.26/

#7 lore raw

Re: Reiser4 needs more testers

In reply to Hans Reiser
Quoted 1 line · Hans Reiser

We have one NFS related bug remaining, [snipped]

Is this the NFS bug you have in mind, or a different one?

I get "Input/output error" when I try to ls the contents of a directory on an nfs exported reiser4 system from the client system.

spc2 is the nfs server, running kernel 2.6.5-rc2 with latest reiser4 patch. spc0 is the nfs client, running kernel 2.6.5-rc3-mm1 with nfs-utils 1.0.6.

[steven@spc2 r4_testing]$ df -T

Filesystem    Type   1K-blocks      Used Available Use% Mounted on
/dev/sda1     ext3      798476    156840    601076  21% /
/dev/sda8     ext3     7147260   5570304   1213892  83% /home
none         tmpfs      257456         0    257456   0% /dev/shm
/dev/sda6     ext3     4024188   1874192   1945572  50% /usr
/dev/sda7     ext3      806368    130220    635184  18% /var

/dev/sda9 reiser4 3930680 241556 3689124 7% /r4_testing

[steven@spc2 steven]$ cat /etc/exports
/home/steven/kernels 128.165.67.190(rw)
/r4_testing 128.165.67.190(rw)

[steven@spc2 r4_testing]$ pwd
/r4_testing
[steven@spc2 r4_testing]$ ls
linux-2.6.5-rc2-r4

I lightly tested the reiser4 filesystem by copying the kernel tree from the /home ext3 system and running one compile.

On the client system, I mounted a few nfs systems:

[steven@spc0 steven]$ df -T

Filesystem    Type    Size  Used Avail Use% Mounted on
/dev/hda7     ext3    471M  111M  337M  25% /
/dev/hda11    ext3     26G   15G   11G  59% /home
/dev/hda1     vfat    2.0G  1.9G  135M  94% /mnt/windows
/dev/hda9     ext3    3.9G  1.8G  2.0G  47% /usr
/dev/hda10    ext3    580M  118M  432M  22% /var
/dev/hda5     vfat    1.9G  920M  963M  49% /win_d
/dev/hda6     vfat    1.8G  1.3G  511M  72% /win_e
spc5:/share/from_spc0
               nfs     34G   25G  8.7G  74% /spc5_backup
spc2:/r4_testing
               nfs    3.8G  236M  3.6G   7% /home/steven/r4_testing
spc2:/home/steven/kernels
               nfs    6.9G  5.4G  1.2G  83% /home/steven/spc2_kernels

[steven@spc0 steven]$ uname -r
2.6.5-rc3-mm1

The ext3 system works just fine:

[steven@spc0 steven]$ ls /home/steven/spc2_kernels

2.6.0-test8-mm1   fs/                              linux-2.6.2-r4/       linux-2.6.5-rc2-r4/  r4/
2.6.0-test9-mm1   getit                            linux-2.6.2-rc2-mm1/  patch-2.6.2-rc2      r4.tar.gz
2.6.2-rc2-mm1     kexec-260test8-mm1.patch         linux-2.6.2.tar.bz2   patch-2.6.5-rc2      reiser4/
2.6.4-mm1         kexec-260test9-mm1-update.patch  linux-2.6.4-mm1/      patch-fixk           reiser4.tar.gz
4g4g_kexec.patch  kexec-260test9.patch             linux-2.6.4.tar.bz2   patch-nikita         thediff
all.diff          linux-2.6.1.tar.bz2              linux-2.6.4-vanilla/  patch-nikita2

But the reiser4 system does not:

[steven@spc0 steven]$ ls /home/steven/r4_testing
ls: /home/steven/r4_testing/linux-2.6.5-rc2-r4: Input/output error

Steven

#8 lore raw

Re: Reiser4 needs more testers

Yes, it is a readdir related NFS bug. I should have added that it is serious enough to make NFS over reiser4 a bad idea.

Hans

Quoted 79 lines · Steven Cole
Quoted 3 lines · Hans Reiser

We have one NFS related bug remaining, [snipped]

Is this the NFS bug you have in mind, or a different one?

I get "Input/output error" when I try to ls the contents of a directory on an nfs exported reiser4 system from the client system.

spc2 is the nfs server, running kernel 2.6.5-rc2 with latest reiser4 patch. spc0 is the nfs client, running kernel 2.6.5-rc3-mm1 with nfs-utils 1.0.6.

[steven@spc2 r4_testing]$ df -T

Filesystem    Type   1K-blocks      Used Available Use% Mounted on
/dev/sda1     ext3      798476    156840    601076  21% /
/dev/sda8     ext3     7147260   5570304   1213892  83% /home
none         tmpfs      257456         0    257456   0% /dev/shm
/dev/sda6     ext3     4024188   1874192   1945572  50% /usr
/dev/sda7     ext3      806368    130220    635184  18% /var

/dev/sda9 reiser4 3930680 241556 3689124 7% /r4_testing

[steven@spc2 steven]$ cat /etc/exports
/home/steven/kernels 128.165.67.190(rw)
/r4_testing 128.165.67.190(rw)

[steven@spc2 r4_testing]$ pwd
/r4_testing
[steven@spc2 r4_testing]$ ls
linux-2.6.5-rc2-r4

I lightly tested the reiser4 filesystem by copying the kernel tree from the /home ext3 system and running one compile.

On the client system, I mounted a few nfs systems:

[steven@spc0 steven]$ df -T

Filesystem    Type    Size  Used Avail Use% Mounted on
/dev/hda7     ext3    471M  111M  337M  25% /
/dev/hda11    ext3     26G   15G   11G  59% /home
/dev/hda1     vfat    2.0G  1.9G  135M  94% /mnt/windows
/dev/hda9     ext3    3.9G  1.8G  2.0G  47% /usr
/dev/hda10    ext3    580M  118M  432M  22% /var
/dev/hda5     vfat    1.9G  920M  963M  49% /win_d
/dev/hda6     vfat    1.8G  1.3G  511M  72% /win_e
spc5:/share/from_spc0
              nfs     34G   25G  8.7G  74% /spc5_backup
spc2:/r4_testing
              nfs    3.8G  236M  3.6G   7% /home/steven/r4_testing
spc2:/home/steven/kernels
              nfs    6.9G  5.4G  1.2G  83% /home/steven/spc2_kernels

[steven@spc0 steven]$ uname -r
2.6.5-rc3-mm1

The ext3 system works just fine:

[steven@spc0 steven]$ ls /home/steven/spc2_kernels

2.6.0-test8-mm1   fs/                              linux-2.6.2-r4/       linux-2.6.5-rc2-r4/  r4/
2.6.0-test9-mm1   getit                            linux-2.6.2-rc2-mm1/  patch-2.6.2-rc2      r4.tar.gz
2.6.2-rc2-mm1     kexec-260test8-mm1.patch         linux-2.6.2.tar.bz2   patch-2.6.5-rc2      reiser4/
2.6.4-mm1         kexec-260test9-mm1-update.patch  linux-2.6.4-mm1/      patch-fixk           reiser4.tar.gz
4g4g_kexec.patch  kexec-260test9.patch             linux-2.6.4.tar.bz2   patch-nikita         thediff
all.diff          linux-2.6.1.tar.bz2              linux-2.6.4-vanilla/  patch-nikita2

But the reiser4 system does not:

[steven@spc0 steven]$ ls /home/steven/r4_testing
ls: /home/steven/r4_testing/linux-2.6.5-rc2-r4: Input/output error

Steven

Signature · 1 line
Hans
#9 lore raw
In reply to Nikita Danilov
Quoted 8 lines · Nikita Danilov

Hello,

new reiser4 snapshot against 2.6.5-rc2 is available at

http://www.namesys.com/snapshots/2004.03.26/

It is mainly bug-fixing release. See READ.ME for the list of fixes and caveats.

A definition of fibration:
http://mathworld.wolfram.com/Fibration.html

I'm going to have to study math for about a year before I understand all that, I think.

It's a good thing we won't have to understand "fiber bundles", "paracompact topological space" and the "homotopy lifting property" to USE Reiser4.

*grin*

If I missed the discussion or a web page, I am sorry. But could someone post a quick explanation or pointer to one about this fibration plugin? What does it do and what effects will it have?

Signature · 2 lines
Jonathan Briggs
[email protected]
#10 lore raw
Quoted 5 lines · Jonathan Briggs
Quoted 4 lines · Nikita Danilov

Hello,

new reiser4 snapshot against 2.6.5-rc2 is available at

Quoted 7 lines
Quoted 4 lines

http://www.namesys.com/snapshots/2004.03.26/

It is mainly bug-fixing release. See READ.ME for the list of fixes and caveats.

A definition of fibration:
http://mathworld.wolfram.com/Fibration.html

Reiser4 plugin was named this was due to some (arguably vague) similarity with mathematical fibrations.

Quoted 9 lines

I'm going to have to study math for about a year before I understand all that, I think.

It's a good thing we won't have to understand "fiber bundles", "paracompact topological space" and the "homotopy lifting property" to USE Reiser4.

*grin*

Why, of course one has to understand it. Reiser4 refuses to mount unless supplied with the homotopy group of the tangent bundle of hard drive, for sure.

Quoted 4 lines

If I missed the discussion or a web page, I am sorry. But could someone post a quick explanation or pointer to one about this fibration plugin? What does it do and what effects will it have?

Fibration plugin affects how disk blocks are allocated for the files within the same directory. Basically, in reiser4 all file system data and meta-data (except for allocator bitmaps) are stored in a single balanced tree. Every piece of information in the file system (byte of file data, on-disk inode, directory entry containing file name, etc.) has a key that allows to locate this information in the tree. This imposes natural order on all file system data (because keys are just large integers, and can be compared).

Block allocator tries to allocate blocks in a parent-first tree order. This means, that things with close keys have chances to be close to each other on a disk. This leads to the main high-level mechanism that reiser4 uses to control disk layout: through key assignment.

In particular fibration plugin is called when new name is inserted into a directory, and, based on a name, selects some (otherwise unused) 7 bits in a key of directory entry. This allows to "slice" directory content into "fibers", hence the name.

For example, one possible implementation is to place .o files in one fiber and all others in another. This significantly speeds compilations up, because .o files are created close to each other and don't interfere with sources. Fibrations, and well as other plugins, can be set per-object, see http://www.namesys.com/v4/pseudo.html for details.

Quoted 5 lines
Signature · 3 lines
-- 
Jonathan Briggs
[email protected]

Nikita.

#11 lore raw

Nikita, would you confirm that the default fibration is to sort all files with '.' as the penultimate character by their last character first and then by the rest of the name in the usual lexicographic order? (If that is not the default, then please make it the default.)

Also, please note that the URL you supplied does not yet describe the fibration plugin and its available settings. Please correct that on Monday.

Signature · 1 line
Hans
#12 lore raw
Quoted 4 lines · Hans Reiser

Nikita, would you confirm that the default fibration is to sort all files with '.' as the penultimate character by their last character first and then by the rest of the name in the usual lexicographic order? (If that is not the default, then please make it the default.)

Wouldn't it be slightly beneficial to expand this policy to not just look for extensions matching .*\.(?) but .*\.(.*) as well?

-- Cyrille, needing some time to dive back into reiser4 testing

#13 lore raw
Quoted 4 lines · Cyrille Chepelov

Le Fri, Mar 26, 2004, Ã 09:11:00AM -0800, Hans Reiser a écrit:

Quoted 11 lines
Quoted 3 lines

Nikita, would you confirm that the default fibration is to sort all files with '.' as the penultimate character by their last character first and then by the rest of the name in the usual lexicographic order? (If that is not the default, then please make it the default.)

Wouldn't it be slightly beneficial to expand this policy to not just look for extensions matching .*\.(?) but .*\.(.*) as well?

-- Cyrille, needing some time to dive back into reiser4 testing

I thought about that, but other than compiles I don't know of an application that would benefit from this performance wise, and compiles usually have . as the penultimate character in their fileset filenames. I would actually much like to get feedback from others on this.

Signature · 1 line
Hans
#14 lore raw
Quoted 7 lines · Hans Reiser
Quoted 2 lines

Wouldn't it be slightly beneficial to expand this policy to not just look for extensions matching .*\.(?) but .*\.(.*) as well?

I thought about that, but other than compiles I don't know of an application that would benefit from this performance wise, and compiles usually have . as the penultimate character in their fileset filenames. I would actually much like to get feedback from others on this.

I see. A little question on the importanceness of solving this right up
front: is it possible to change the fibration policy of a mounted and active file system? Does such a change affect the repacker (as in: swapping the fibration policy causes the repacker to start shuffling things around according to the new policy)? If that's the case, then I guess there's some wiggle room for imperfection...

(in fact, on second thought, my request wouldn't necessarily cause the best performance to happen. Norton Defrag used to have a "sort by type" option ca. 1994, and I remember that on my PC, it didn't give the best performance).

Just out of curiosity, is it possible to create a fibration policy based on atime/mtime/ctime?

-- Cyrille

#15 lore raw

Hello

Quoted 6 lines · Cyrille Chepelov
Quoted 2 lines · Hans Reiser
Quoted 2 lines

Wouldn't it be slightly beneficial to expand this policy to not just look for extensions matching .*\.(?) but .*\.(.*) as well?

Quoted 8 lines
Quoted 4 lines

I thought about that, but other than compiles I don't know of an application that would benefit from this performance wise, and compiles usually have . as the penultimate character in their fileset filenames. I would actually much like to get feedback from others on this.

I see. A little question on the importanceness of solving this right up
front: is it possible to change the fibration policy of a mounted and active file system?

Changing fibration policy applies only to empty directory

Quoted 3 lines

Does such a change affect the repacker (as in: swapping the
fibration policy causes the repacker to start shuffling things around according to the new policy)?

repacker does not get affected

Quoted 12 lines

If that's the case, then I guess there's some
wiggle room for imperfection...

(in fact, on second thought, my request wouldn't necessarily cause the best performance to happen. Norton Defrag used to have a "sort by type" option ca. 1994, and I remember that on my PC, it didn't give the best performance).

Just out of curiosity, is it possible to create a fibration policy based on atime/mtime/ctime?

-- Cyrille

#16 lore raw
In reply to Nikita Danilov

ND> Why, of course one has to understand it. Reiser4 refuses to mount unless ND> supplied with the homotopy group of the tangent bundle of hard drive, ND> for sure.
Oh man! Is reiser4 a part of some alien or future technology? From what I read and hear it seems so new, fresh and advanced, it seems _obvious_ :-) this code has been inspired by aliens or time travellers from the future. Or maybe it's just me being not so knowledgible as reiser4 gurus.

Regards,
Maciej