reiserfs and nfs.

reiserfs and nfs.

[Terje Malmedal]

Hi, 

I'm nfs-exporting a reiserfs filesystem, the problem is that the
inode-number as seen from the client seems to change from time to
time. This confuses a number of programs, for instance Emacs believes
that the file changed under it when this happens.

Is this a known problem? Any known fix? I found a number of old
messages about problems with reiser and nfs, but not this
specifically.

-- 
 - Terje
malmedal@usit.uio.no

Re: reiserfs and nfs.

[Nikita Danilov]

Terje Malmedal writes:
 > 
 > Hi, 
 > 
 > I'm nfs-exporting a reiserfs filesystem, the problem is that the
 > inode-number as seen from the client seems to change from time to
 > time. This confuses a number of programs, for instance Emacs believes
 > that the file changed under it when this happens.

This problem is not known. What version of reiserfs (3.5 or 3.6) and
kernel are you using? Is there any way to reproduce this problem?

 > 
 > Is this a known problem? Any known fix? I found a number of old
 > messages about problems with reiser and nfs, but not this
 > specifically.
 > 
 > -- 
 >  - Terje
 > malmedal@usit.uio.no

Nikita.

Re: reiserfs and nfs.

[Terje Malmedal]

[Nikita Danilov]
> Terje Malmedal writes:
>> 
>> Hi, 
>> 
>> I'm nfs-exporting a reiserfs filesystem, the problem is that the
>> inode-number as seen from the client seems to change from time to
>> time. This confuses a number of programs, for instance Emacs believes
>> that the file changed under it when this happens.

> This problem is not known. What version of reiserfs (3.5 or 3.6) and
> kernel are you using? Is there any way to reproduce this problem?

On the NFS-server I'm using RedHat kernel 2.4.18-10smp, according to
dmesg I'm using version 3.6 of reiser: 
reiserfs: checking transaction log (device 3a:00) ...
Using r5 hash to sort names
ReiserFS version 3.6.25

(The block device is LVM by the way)

Clients are both Linux and Solaris. 

To reproduce do something like: 
$ ls -li RMAIL
 387607 -rw-------    1 tm       4050     36635726 Nov 20 16:15 RMAIL

on the nfs-client, (obviously on a NFS-file) and wait a bit, varying
from a few minutes to many hours and the inode-number will be
different. I'm guessing the period you have to wait is dependent on
how much else is going on on the NFS-server.

I have no reason to suspect any data-corruption, md5 of file as well
as atime, ctime and mtime are unchanged.

-- 
 - Terje
malmedal@usit.uio.no

Re: reiserfs and nfs.

[Nikita Danilov]

Terje Malmedal writes:
 > 
 > [Nikita Danilov]
 > > Terje Malmedal writes:
 > >> 

[...]

 > 
 > To reproduce do something like: 
 > $ ls -li RMAIL
 >  387607 -rw-------    1 tm       4050     36635726 Nov 20 16:15 RMAIL

Is RMAIL inode number constant on the server?

Nikita.

Re: reiserfs and nfs.

[Terje Malmedal]

[Nikita Danilov]
> Terje Malmedal writes:
>> 
>> [Nikita Danilov]
>> > Terje Malmedal writes:
>> >> 

> [...]

>> 
>> To reproduce do something like: 
>> $ ls -li RMAIL
>> 387607 -rw-------    1 tm       4050     36635726 Nov 20 16:15 RMAIL

> Is RMAIL inode number constant on the server?

No, it is always the same as the number seen over NFS, but I've not
seen this problem when I've tried running emacs on the nfs-server
itself.

-- 
 - Terje
malmedal@usit.uio.no

Re: reiserfs and nfs.

[Nikita Danilov]

[lkml trimmed from CC list.]

Terje Malmedal writes:
 > 
 > [Nikita Danilov]
 > > Terje Malmedal writes:
 > >> 
 > >> [Nikita Danilov]
 > >> > Terje Malmedal writes:
 > >> >> 
 > 
 > > [...]
 > 
 > >> 
 > >> To reproduce do something like: 
 > >> $ ls -li RMAIL
 > >> 387607 -rw-------    1 tm       4050     36635726 Nov 20 16:15 RMAIL
 > 
 > > Is RMAIL inode number constant on the server?
 > 
 > No, it is always the same as the number seen over NFS, but I've not
 > seen this problem when I've tried running emacs on the nfs-server
 > itself.

I see following possibilities:

1. some weird mail locking protocol (is RMAIL your mail spool file?)

2. some EMACS/Rmail feature. Are you running identical Emacs
configurations on the client and server?

 > 
 > -- 
 >  - Terje
 > malmedal@usit.uio.no

Nikita.

Re: reiserfs and nfs.

[Terje Malmedal]

[Nikita Danilov]
> [lkml trimmed from CC list.]
> Terje Malmedal writes:
>> 
>> [Nikita Danilov]
>> > Terje Malmedal writes:
>> >> 
>> >> [Nikita Danilov]
>> >> > Terje Malmedal writes:
>> >> >> 
>> 
>> > [...]
>> 
>> >> 
>> >> To reproduce do something like: 
>> >> $ ls -li RMAIL
>> >> 387607 -rw-------    1 tm       4050     36635726 Nov 20 16:15 RMAIL
>> 
>> > Is RMAIL inode number constant on the server?
>> 
>> No, it is always the same as the number seen over NFS, but I've not
>> seen this problem when I've tried running emacs on the nfs-server
>> itself.

> I see following possibilities:

> 1. some weird mail locking protocol (is RMAIL your mail spool file?)

No, emacs gets the mail via POP3 and saves it in that file.

This also happens with normal files being edited in emacs. 

> 2. some EMACS/Rmail feature. Are you running identical Emacs
> configurations on the client and server?

Yes, both are RedHat 7.3 standard emacs, also this did not happen when
the nfs-server was a solaris-box.

-- 
 - Terje
malmedal@usit.uio.no

Re: reiserfs and nfs.

[Anders Widman]

> [Nikita Danilov]
>> [lkml trimmed from CC list.]
>> Terje Malmedal writes:
>>> 
>>> [Nikita Danilov]
>>> > Terje Malmedal writes:
>>> >> 
>>> >> [Nikita Danilov]
>>> >> > Terje Malmedal writes:
>>> >> >> 
>>> 
>>> > [...]
>>> 
>>> >> 
>>> >> To reproduce do something like: 
>>> >> $ ls -li RMAIL
>>> >> 387607 -rw-------    1 tm       4050     36635726 Nov 20 16:15 RMAIL
>>> 
>>> > Is RMAIL inode number constant on the server?
>>> 
>>> No, it is always the same as the number seen over NFS, but I've not
>>> seen this problem when I've tried running emacs on the nfs-server
>>> itself.

>> I see following possibilities:

>> 1. some weird mail locking protocol (is RMAIL your mail spool file?)

> No, emacs gets the mail via POP3 and saves it in that file.

> This also happens with normal files being edited in emacs. 

>> 2. some EMACS/Rmail feature. Are you running identical Emacs
>> configurations on the client and server?

> Yes, both are RedHat 7.3 standard emacs, also this did not happen when
> the nfs-server was a solaris-box.



I  had  problems with NFS under Redhat 7.2 and 7.3. It works fine with
8.0 and other distributions such as LFS and Mandrake 9.

- Anders

Any information on S+tree?

[bmoon]

Hello,

I thought Reiserfs uses B+tree, but 
Reiserfs is using S+tree according to the source code.
Where can I get any information or document on this S+tree?

Thanks,

Bo

Re: Any information on S+tree?

[Yury Umanets]

bmoon wrote:

>Hello,
>
>I thought Reiserfs uses B+tree, but 
>Reiserfs is using S+tree according to the source code.
>Where can I get any information or document on this S+tree?
>
>Thanks,
>
>Bo
>
>
>  
>
Read the source. Take a lot onto function search_by_key.

-- 
Yury Umanets

Re: Any information on S+tree?

[Yury Umanets]

Yury Umanets wrote:

> bmoon wrote:
>
>> Hello,
>>
>> I thought Reiserfs uses B+tree, but Reiserfs is using S+tree 
>> according to the source code.
>> Where can I get any information or document on this S+tree?
>>
>> Thanks,
>>
>> Bo
>>
>>
>>  
>>
> Read the source. Take a lot onto function search_by_key.
>
:) I have meant take a look

-- 
Yury Umanets

Re: Any information on S+tree?

[Hans Reiser]

bmoon wrote:

>Hello,
>
>I thought Reiserfs uses B+tree, but 
>Reiserfs is using S+tree according to the source code.
>Where can I get any information or document on this S+tree?
>
>Thanks,
>
>Bo
>
>
>  
>
S+tree is an old name for what we use.  You might look at our 
www.namesys.com/v4/v4p1.html and www.namesys.com/v4/v4p2.html and 
www.namesys.com/v4/v4.html documentation for our latest trees.

-- 
Hans

any tips on optimising samba on reiserfs?

[darren]

Hi all,

I have been trying to setup my reiserfs/samba CIFS file server on my
LAN.

Although all my benchmarks have been showing good local read write
performance, I cannot seem to get the most of it over the network with
samba.

Anyone got any tips on how I can do this?

p.s. I have already done all the obvious optimizations like TCP_NODELAY
and using SO_SNBUF=16384 blah blah..

Re: any tips on optimising samba on reiserfs?

[Hans Reiser]

darren wrote:

>Hi all,
>
>I have been trying to setup my reiserfs/samba CIFS file server on my
>LAN.
>
>Although all my benchmarks have been showing good local read write
>performance, I cannot seem to get the most of it over the network with
>samba.
>
>Anyone got any tips on how I can do this?
>
>p.s. I have already done all the obvious optimizations like TCP_NODELAY
>and using SO_SNBUF=16384 blah blah..
>
>
>
>  
>
This is an area where our team is ignorant due to lack of experience, 
but I am very interested in learning from you and other users about it, 
and we are willing to tweak as needed.  Perhaps we should run netbench 
or some such to replicate your situation?  (We don't use much Windows at 
Namesys....)

-- 
Hans