"df -i" support

"df -i" support

[Miguel]

Hi
We have several machines running as MTA's using reiser for the mail
queue, we usually check the number of files (well an approximation)
using the output of "df -i" command on our monitorization plugins this
may be not very acurate but is very fast and didn't affect the MTA's
work.

With reiser the output of df -i is no longer usable, we changed this to
"find | wc -l" variations but it hurts the performance.

There's a command to count the number of leafs (that's the name?) of a
tree from a reiser partition? or other orientative value?.(we mount
with notail)

Tnx

-- 
La resistencia es fútil todos seréis asimilados

Re: "df -i" support

[Vladimir Saveliev]

Hello

On Thu, 2004-05-13 at 11:34, Miguel wrote:
> Hi
> We have several machines running as MTA's using reiser for the mail
> queue, we usually check the number of files (well an approximation)
> using the output of "df -i" command on our monitorization plugins this
> may be not very acurate but is very fast and didn't affect the MTA's
> work.
> 
> With reiser the output of df -i is no longer usable,

IIRC, reiserfs always reported -1 as a total number of inodes and a
number of free inodes.
At whihc kernel it was useable for you?

>  we changed this to
> "find | wc -l" variations but it hurts the performance.
> 
> There's a command to count the number of leafs (that's the name?) of a
> tree from a reiser partition? or other orientative value?.(we mount
> with notail)
> 
> Tnx

Re: "df -i" support

[Miguel]

On Thu, 13 May 2004 13:19:06 +0400
Vladimir Saveliev <vs@namesys.com> wrote:

> Hello
> 
> On Thu, 2004-05-13 at 11:34, Miguel wrote:
> > Hi
> > We have several machines running as MTA's using reiser for the mail
> > queue, we usually check the number of files (well an approximation)
> > using the output of "df -i" command on our monitorization plugins
> > this may be not very acurate but is very fast and didn't affect the
> > MTA's work.
> > 
> > With reiser the output of df -i is no longer usable,
> 
> IIRC, reiserfs always reported -1 as a total number of inodes and a
> number of free inodes.
> At whihc kernel it was useable for you?

mmmh seems that my question was not very clear

We are migrating our mtas from ext3 to reiserfs and the monitorization
(a plugin) relies on the output of df -i, this output is very diferent
between those filesystems, because reiserfs always return -1 as a total
number of inodes.

I want to know if there's another command to get similar info from
reiserfs partitions since reiserfs doesn't use/has inodes (I think that
you call this leafs).


IFAIK you implement some kind of trickery related to inodes to make NFS
work in the past, this is correct? You can easily map (virtual)inodes to
whatever internal estructure reiserfs has?.


> 
> >  we changed this to
> > "find | wc -l" variations but it hurts the performance.
> > 
> > There's a command to count the number of leafs (that's the name?) of
> > a tree from a reiser partition? or other orientative value?.(we
> > mount with notail)
> > 
> > Tnx
> 


-- 
La resistencia es fútil todos seréis asimilados

Re: "df -i" support

[Valdis.Kletnieks]

[-- Attachment #1: Type: text/plain, Size: 812 bytes --]

On Fri, 14 May 2004 17:32:27 +0200, Miguel said:

> We are migrating our mtas from ext3 to reiserfs and the monitorization
> (a plugin) relies on the output of df -i, this output is very diferent
> between those filesystems, because reiserfs always return -1 as a total
> number of inodes.

The only reason you care about df -i is because on ext[23] and similar,
it is possible to run out of inodes before you run out of files.  Since
reiserfs allocates inodes as needed, you can't run out of them unless you
run out of space - in which case your monitoring system should be flagging
that.

So just fix your code to say '-1 means unlimited', or if that's too complicated
make it say:

	if (value == -1) { value = 2147483648; }

which won't be right but should make your scripts act correctly.

[-- Attachment #2: Type: application/pgp-signature, Size: 226 bytes --]

Re: "df -i" support

[Miguel]

On Fri, 14 May 2004 12:55:53 -0400
Valdis.Kletnieks@vt.edu wrote:

> On Fri, 14 May 2004 17:32:27 +0200, Miguel said:
> 
> > We are migrating our mtas from ext3 to reiserfs and the
> > monitorization(a plugin) relies on the output of df -i, this output
> > is very diferent between those filesystems, because reiserfs always
> > return -1 as a total number of inodes.
> 
> The only reason you care about df -i is because on ext[23] and
> similar, it is possible to run out of inodes before you run out of
> files.  Since reiserfs allocates inodes as needed, you can't run out
> of them unless you run out of space - in which case your monitoring
> system should be flagging that.

No, we check this value to know how many files are in that
partition (well an aproximation). I hope that this is true for ext2/3,
isn't?

> 
> So just fix your code to say '-1 means unlimited', or if that's too
> complicated make it say:
> 
> 	if (value == -1) { value = 2147483648; }
> 
> which won't be right but should make your scripts act correctly.

df -i on a 2.4.x kernel already returns (2^32 - 1) free inodes.
probably kernel return -1 but df "get" this as a non-signed value.

Here's an example sda1 is an ext3 partition an sdb1 is reiserfs
formatted:


Filesystem            Inodes   IUsed   IFree IUse% Mounted on
/dev/sda1            4480448  132216 4348232    3% /
/dev/sdb1            4294967295       0 4294967295    0%
/var/spool/postfix

scripts return about 132000 files in / and 0 files in the other
filesystem


-- 
La resistencia es fútil todos seréis asimilados

Re: "df -i" support

[Vladimir Saveliev]

Hello

On Fri, 2004-05-14 at 21:40, Miguel wrote:
> On Fri, 14 May 2004 12:55:53 -0400
> Valdis.Kletnieks@vt.edu wrote:
> 
> > On Fri, 14 May 2004 17:32:27 +0200, Miguel said:
> > 
> > > We are migrating our mtas from ext3 to reiserfs and the
> > > monitorization(a plugin) relies on the output of df -i, this output
> > > is very diferent between those filesystems, because reiserfs always
> > > return -1 as a total number of inodes.
> > 
> > The only reason you care about df -i is because on ext[23] and
> > similar, it is possible to run out of inodes before you run out of
> > files.  Since reiserfs allocates inodes as needed, you can't run out
> > of them unless you run out of space - in which case your monitoring
> > system should be flagging that.
> 
> No, we check this value to know how many files are in that
> partition (well an aproximation).

Well, it is possible to calculate approximately how many files are on a
filesystem

>  I hope that this is true for ext2/3,
> isn't?

> > 
> > So just fix your code to say '-1 means unlimited', or if that's too
> > complicated make it say:
> > 
> > 	if (value == -1) { value = 2147483648; }
> > 
> > which won't be right but should make your scripts act correctly.
> 
> df -i on a 2.4.x kernel already returns (2^32 - 1) free inodes.
> probably kernel return -1 but df "get" this as a non-signed value.
> 
> Here's an example sda1 is an ext3 partition an sdb1 is reiserfs
> formatted:
> 
> 
> Filesystem            Inodes   IUsed   IFree IUse% Mounted on
> /dev/sda1            4480448  132216 4348232    3% /
> /dev/sdb1            4294967295       0 4294967295    0%
> /var/spool/postfix
> 
> scripts return about 132000 files in / and 0 files in the other
> filesystem
> 

Re: "df -i" support

[Miguel]

On Fri, 14 May 2004 21:48:22 +0400
Vladimir Saveliev <vs@namesys.com> wrote:

> Hello
> 
> On Fri, 2004-05-14 at 21:40, Miguel wrote:
> > On Fri, 14 May 2004 12:55:53 -0400
> > Valdis.Kletnieks@vt.edu wrote:
> > 
> > > On Fri, 14 May 2004 17:32:27 +0200, Miguel said:
> > > 
> > > > We are migrating our mtas from ext3 to reiserfs and the
> > > > monitorization(a plugin) relies on the output of df -i, this
> > > > output is very diferent between those filesystems, because
> > > > reiserfs always return -1 as a total number of inodes.
> > > 
> > > The only reason you care about df -i is because on ext[23] and
> > > similar, it is possible to run out of inodes before you run out of
> > > files.  Since reiserfs allocates inodes as needed, you can't run
> > > out of them unless you run out of space - in which case your
> > > monitoring system should be flagging that.
> > 
> > No, we check this value to know how many files are in that
> > partition (well an aproximation).
> 
> Well, it is possible to calculate approximately how many files are on
> a filesystem

Yes?, how should I calculate this?

> 
> >  I hope that this is true for ext2/3,
> > isn't?
> 
> > > 
> > > So just fix your code to say '-1 means unlimited', or if that's
> > > too complicated make it say:
> > > 
> > > 	if (value == -1) { value = 2147483648; }
> > > 
> > > which won't be right but should make your scripts act correctly.
> > 
> > df -i on a 2.4.x kernel already returns (2^32 - 1) free inodes.
> > probably kernel return -1 but df "get" this as a non-signed value.
> > 
> > Here's an example sda1 is an ext3 partition an sdb1 is reiserfs
> > formatted:
> > 
> > 
> > Filesystem            Inodes   IUsed   IFree IUse% Mounted on
> > /dev/sda1            4480448  132216 4348232    3% /
> > /dev/sdb1            4294967295       0 4294967295    0%
> > /var/spool/postfix
> > 
> > scripts return about 132000 files in / and 0 files in the other
> > filesystem
> > 
> 


-- 
La resistencia es fútil todos seréis asimilados

Re: "df -i" support

[Jonathan Briggs]

On Fri, 2004-05-14 at 11:40, Miguel wrote:
> On Fri, 14 May 2004 12:55:53 -0400
> Valdis.Kletnieks@vt.edu wrote:
> 
> > On Fri, 14 May 2004 17:32:27 +0200, Miguel said:
> > 
> > > We are migrating our mtas from ext3 to reiserfs and the
> > > monitorization(a plugin) relies on the output of df -i, this output
> > > is very diferent between those filesystems, because reiserfs always
> > > return -1 as a total number of inodes.
> > 
> > The only reason you care about df -i is because on ext[23] and
> > similar, it is possible to run out of inodes before you run out of
> > files.  Since reiserfs allocates inodes as needed, you can't run out
> > of them unless you run out of space - in which case your monitoring
> > system should be flagging that.
> 
> No, we check this value to know how many files are in that
> partition (well an aproximation). I hope that this is true for ext2/3,
> isn't?
> 

Would this help you?

# cat /proc/fs/reiserfs/md1/oidmap

At the end of oidmap is a line like this:
total:  46 [46/972] used: 176147 [exact]

I think the used: 176147 is pretty close to what you want.  The number
goes up and down when I create or remove files.  When I ran 'find /
-xdev | wc -l' I got 174425.  I'm not sure where the extra 1,000+ comes
from.  

You would need to have CONFIG_REISERFS_PROC_INFO set in your kernel, and
of course you would need to replace md1 with whatever device you were
using.
-- 
Jonathan Briggs
jbriggs@esoft.com

Re: "df -i" support

[Nikita Danilov]

Jonathan Briggs writes:
 > On Fri, 2004-05-14 at 11:40, Miguel wrote:
 > > On Fri, 14 May 2004 12:55:53 -0400
 > > Valdis.Kletnieks@vt.edu wrote:
 > > 
 > > > On Fri, 14 May 2004 17:32:27 +0200, Miguel said:
 > > > 
 > > > > We are migrating our mtas from ext3 to reiserfs and the
 > > > > monitorization(a plugin) relies on the output of df -i, this output
 > > > > is very diferent between those filesystems, because reiserfs always
 > > > > return -1 as a total number of inodes.
 > > > 
 > > > The only reason you care about df -i is because on ext[23] and
 > > > similar, it is possible to run out of inodes before you run out of
 > > > files.  Since reiserfs allocates inodes as needed, you can't run out
 > > > of them unless you run out of space - in which case your monitoring
 > > > system should be flagging that.
 > > 
 > > No, we check this value to know how many files are in that
 > > partition (well an aproximation). I hope that this is true for ext2/3,
 > > isn't?
 > > 
 > 
 > Would this help you?
 > 
 > # cat /proc/fs/reiserfs/md1/oidmap
 > 
 > At the end of oidmap is a line like this:
 > total:  46 [46/972] used: 176147 [exact]
 > 
 > I think the used: 176147 is pretty close to what you want.  The number

Yes, it is. On the other hand, oidmap can "overflow", then only
estimated number of files will be shown.

 > goes up and down when I create or remove files.  When I ran 'find /
 > -xdev | wc -l' I got 174425.  I'm not sure where the extra 1,000+ comes
 > from.  

"find | wc -l" returns number of _names_ rather than files. I.e.,
hard-linked files are counter more than once.

 > 
 > You would need to have CONFIG_REISERFS_PROC_INFO set in your kernel, and
 > of course you would need to replace md1 with whatever device you were
 > using.
 > -- 
 > Jonathan Briggs
 > jbriggs@esoft.com
 > 

Nikita.

Re: "df -i" support

[Miguel]

On Fri, 14 May 2004 11:50:09 -0600
Jonathan Briggs <jbriggs@esoft.com> wrote:

> On Fri, 2004-05-14 at 11:40, Miguel wrote:
> > On Fri, 14 May 2004 12:55:53 -0400
> > Valdis.Kletnieks@vt.edu wrote:
> > 
> > > On Fri, 14 May 2004 17:32:27 +0200, Miguel said:
> > > 
> > > > We are migrating our mtas from ext3 to reiserfs and the
> > > > monitorization(a plugin) relies on the output of df -i, this
> > > > output is very diferent between those filesystems, because
> > > > reiserfs always return -1 as a total number of inodes.
> > > 
> > > The only reason you care about df -i is because on ext[23] and
> > > similar, it is possible to run out of inodes before you run out of
> > > files.  Since reiserfs allocates inodes as needed, you can't run
> > > out of them unless you run out of space - in which case your
> > > monitoring system should be flagging that.
> > 
> > No, we check this value to know how many files are in that
> > partition (well an aproximation). I hope that this is true for
> > ext2/3, isn't?
> > 
> 
> Would this help you?
> 
> # cat /proc/fs/reiserfs/md1/oidmap
> 
> At the end of oidmap is a line like this:
> total:  46 [46/972] used: 176147 [exact]
> 
> I think the used: 176147 is pretty close to what you want.  The number
> goes up and down when I create or remove files.  When I ran 'find /
> -xdev | wc -l' I got 174425.  I'm not sure where the extra 1,000+
> comes from.  
> 
> You would need to have CONFIG_REISERFS_PROC_INFO set in your kernel,
> and of course you would need to replace md1 with whatever device you
> were using.

I can't set this kind of debuging in a production enviroment but those
are the values that i'm seeking for. There's another way to get this?.

Thanks for this hint.

> -- 
> Jonathan Briggs
> jbriggs@esoft.com
> 


-- 
La resistencia es fútil todos seréis asimilados

Re: "df -i" support

[Nikita Danilov]

Miguel writes:
 > On Fri, 14 May 2004 11:50:09 -0600
 > Jonathan Briggs <jbriggs@esoft.com> wrote:
 > 
 > > On Fri, 2004-05-14 at 11:40, Miguel wrote:
 > > > On Fri, 14 May 2004 12:55:53 -0400
 > > > Valdis.Kletnieks@vt.edu wrote:
 > > > 
 > > > > On Fri, 14 May 2004 17:32:27 +0200, Miguel said:
 > > > > 
 > > > > > We are migrating our mtas from ext3 to reiserfs and the
 > > > > > monitorization(a plugin) relies on the output of df -i, this
 > > > > > output is very diferent between those filesystems, because
 > > > > > reiserfs always return -1 as a total number of inodes.
 > > > > 
 > > > > The only reason you care about df -i is because on ext[23] and
 > > > > similar, it is possible to run out of inodes before you run out of
 > > > > files.  Since reiserfs allocates inodes as needed, you can't run
 > > > > out of them unless you run out of space - in which case your
 > > > > monitoring system should be flagging that.
 > > > 
 > > > No, we check this value to know how many files are in that
 > > > partition (well an aproximation). I hope that this is true for
 > > > ext2/3, isn't?
 > > > 
 > > 
 > > Would this help you?
 > > 
 > > # cat /proc/fs/reiserfs/md1/oidmap
 > > 
 > > At the end of oidmap is a line like this:
 > > total:  46 [46/972] used: 176147 [exact]
 > > 
 > > I think the used: 176147 is pretty close to what you want.  The number
 > > goes up and down when I create or remove files.  When I ran 'find /
 > > -xdev | wc -l' I got 174425.  I'm not sure where the extra 1,000+
 > > comes from.  
 > > 
 > > You would need to have CONFIG_REISERFS_PROC_INFO set in your kernel,
 > > and of course you would need to replace md1 with whatever device you
 > > were using.
 > 
 > I can't set this kind of debuging in a production enviroment but those
 > are the values that i'm seeking for. There's another way to get this?.

This is not debugging. It doesn't affect performance.

 > 
 > Thanks for this hint.
 > 
 > > -- 
 > > Jonathan Briggs
 > > jbriggs@esoft.com
 > > 
 > 

Nikita.

 > 
 > -- 
 > La resistencia es fútil todos seréis asimilados

Re: "df -i" support

[Carlos Carvalho]

The number of files can be obtained from the quota control as well.
It's more complicated to extract the number but gives more detailed
user information.

Re: "df -i" support

[Andreas Dilger]

On May 14, 2004  20:26 +0200, Miguel wrote:
> Jonathan Briggs <jbriggs@esoft.com> wrote:
> > Would this help you?
> > 
> > # cat /proc/fs/reiserfs/md1/oidmap
> > 
> > At the end of oidmap is a line like this:
> > total:  46 [46/972] used: 176147 [exact]
> > 
> > I think the used: 176147 is pretty close to what you want.  The number
> > goes up and down when I create or remove files.  When I ran 'find /
> > -xdev | wc -l' I got 174425.  I'm not sure where the extra 1,000+
> > comes from.  
> > 
> > You would need to have CONFIG_REISERFS_PROC_INFO set in your kernel,
> > and of course you would need to replace md1 with whatever device you
> > were using.
> 
> I can't set this kind of debuging in a production enviroment but those
> are the values that i'm seeking for. There's another way to get this?.

Hmm, why wouldn't reiserfs_statfs() just use that number for the "in use"
inodes in the first place?  Sadly, the statfs() interface is broken in
the sense that it contains "total" inodes and "free" inodes instead of
"in use" inodes, but it shouldn't be difficult to just subtract the
number of in-use inodes from 2^32-1 and stick that into the "free" field
so that when user-space does the reverse it gets the right number.

Cheers, Andreas
--
Andreas Dilger
http://sourceforge.net/projects/ext2resize/
http://www-mddsp.enel.ucalgary.ca/People/adilger/

Re: "df -i" support

[Valdis.Kletnieks]

[-- Attachment #1: Type: text/plain, Size: 608 bytes --]

On Fri, 14 May 2004 12:48:23 MDT, Andreas Dilger said:

> Hmm, why wouldn't reiserfs_statfs() just use that number for the "in use"
> inodes in the first place?  Sadly, the statfs() interface is broken in
> the sense that it contains "total" inodes and "free" inodes instead of
> "in use" inodes, but it shouldn't be difficult to just subtract the
> number of in-use inodes from 2^32-1 and stick that into the "free" field
> so that when user-space does the reverse it gets the right number.

If you get a multi-terabyte filesystem that contains more than 2^32-1 files,
things will get very pear-shaped....


[-- Attachment #2: Type: application/pgp-signature, Size: 226 bytes --]

Re: "df -i" support

[Jeff Mahoney]

Valdis.Kletnieks@vt.edu wrote:
> On Fri, 14 May 2004 12:48:23 MDT, Andreas Dilger said:
> 
> 
>>Hmm, why wouldn't reiserfs_statfs() just use that number for the "in use"
>>inodes in the first place?  Sadly, the statfs() interface is broken in
>>the sense that it contains "total" inodes and "free" inodes instead of
>>"in use" inodes, but it shouldn't be difficult to just subtract the
>>number of in-use inodes from 2^32-1 and stick that into the "free" field
>>so that when user-space does the reverse it gets the right number.
> 
> 
> If you get a multi-terabyte filesystem that contains more than 2^32-1 files,
> things will get very pear-shaped....

If you get a ReiserFS v3 filesystem to contain more than 2^32-1 files, 
you've violated the disk format.

-Jeff

-- 
Jeff Mahoney
SuSE Labs

Re: "df -i" support

[Miguel]

On Fri, 14 May 2004 12:48:23 -0600
Andreas Dilger <adilger@clusterfs.com> wrote:

> On May 14, 2004  20:26 +0200, Miguel wrote:
> > Jonathan Briggs <jbriggs@esoft.com> wrote:
> > > Would this help you?
> > > 
> > > # cat /proc/fs/reiserfs/md1/oidmap
> > > 
> > > At the end of oidmap is a line like this:
> > > total:  46 [46/972] used: 176147 [exact]
> > > 
> > > I think the used: 176147 is pretty close to what you want.  The
> > > number goes up and down when I create or remove files.  When I ran
> > > 'find /-xdev | wc -l' I got 174425.  I'm not sure where the extra
> > > 1,000+ comes from.  
> > > 
> > > You would need to have CONFIG_REISERFS_PROC_INFO set in your
> > > kernel, and of course you would need to replace md1 with whatever
> > > device you were using.
> > 
> > I can't set this kind of debuging in a production enviroment but
> > those are the values that i'm seeking for. There's another way to
> > get this?.
> 
> Hmm, why wouldn't reiserfs_statfs() just use that number for the "in
> use" inodes in the first place?  Sadly, the statfs() interface is
> broken in the sense that it contains "total" inodes and "free" inodes
> instead of"in use" inodes, but it shouldn't be difficult to just
> subtract the number of in-use inodes from 2^32-1 and stick that into
> the "free" field so that when user-space does the reverse it gets the
> right number.

I see your point doing an strace df -i (with a 2.6.6 kernel)
reiserfs_statfs reports those values zeroed:

statfs64("/", 84, {f_type="REISERFS_SUPER_MAGIC", f_bsize=4096,
f_blocks=2560271, f_bfree=497245, f_bavail=497245, f_files=0, f_ffree=0,
f_fsid={0, 0}, f_namelen=255, f_frsize=4096}) = 0

f_files and f_ffree are 0???. I'm trying to do a patch that show those
correctly using show_oidmap() function, but seems that other fs don't
report those values correctly, it's going to be deprecated or something?

> 
> Cheers, Andreas
> --
> Andreas Dilger
> http://sourceforge.net/projects/ext2resize/
> http://www-mddsp.enel.ucalgary.ca/People/adilger/
> 


-- 
La resistencia es fútil todos seréis asimilados

Re: "df -i" support

[Nikita Danilov]

Miguel writes:

[...]

 > 
 > I see your point doing an strace df -i (with a 2.6.6 kernel)
 > reiserfs_statfs reports those values zeroed:
 > 
 > statfs64("/", 84, {f_type="REISERFS_SUPER_MAGIC", f_bsize=4096,
 > f_blocks=2560271, f_bfree=497245, f_bavail=497245, f_files=0, f_ffree=0,
 > f_fsid={0, 0}, f_namelen=255, f_frsize=4096}) = 0
 > 
 > f_files and f_ffree are 0???. I'm trying to do a patch that show those
 > correctly using show_oidmap() function, but seems that other fs don't
 > report those values correctly, it's going to be deprecated or something?

From statfs(2) man page:
----------------------------------------------------------------------

       Fields that are undefined for a particular file system are
       set to 0.

----------------------------------------------------------------------

Fields like
                 long    f_files;    /* total file nodes in file system */
                 long    f_ffree;    /* free file nodes in fs */

are undefined for file systems that, like reiserfs, have no predefined
number of available inodes.

Nikita.

Re: "df -i" support

[Miguel]

On Thu, 20 May 2004 14:35:11 +0400
Nikita Danilov <Nikita@Namesys.COM> wrote:

> Miguel writes:
> 
> [...]
> 
>  > 
>  > I see your point doing an strace df -i (with a 2.6.6 kernel)
>  > reiserfs_statfs reports those values zeroed:
>  > 
>  > statfs64("/", 84, {f_type="REISERFS_SUPER_MAGIC", f_bsize=4096,
>  > f_blocks=2560271, f_bfree=497245, f_bavail=497245, f_files=0,
>  > f_ffree=0, f_fsid={0, 0}, f_namelen=255, f_frsize=4096}) = 0
>  > 
>  > f_files and f_ffree are 0???. I'm trying to do a patch that show
>  > those correctly using show_oidmap() function, but seems that other
>  > fs don't report those values correctly, it's going to be deprecated
>  > or something?
> 
> From statfs(2) man page:
> ----------------------------------------------------------------------
> 
>        Fields that are undefined for a particular file system are
>        set to 0.
> 
> ----------------------------------------------------------------------
> 
> Fields like
>                  long    f_files;    /* total file nodes in file
>                  system */ long    f_ffree;    /* free file nodes in
>                  fs */
> 
> are undefined for file systems that, like reiserfs, have no predefined
> number of available inodes.

Out of curiosity, which is the diference between a file and a file node
in this context? why f_files != (2^32)-1 for 3.6 disk format? in other
words a reiserfs partition can hold more than 4G of dynamically
allocated inodes?

tnx

Re: "df -i" support

[Miguel]

[-- Attachment #1: Type: text/plain, Size: 2024 bytes --]

On Fri, 14 May 2004 11:50:09 -0600
Jonathan Briggs <jbriggs@esoft.com> wrote:

> On Fri, 2004-05-14 at 11:40, Miguel wrote:
> > On Fri, 14 May 2004 12:55:53 -0400
> > Valdis.Kletnieks@vt.edu wrote:
> > 
> > > On Fri, 14 May 2004 17:32:27 +0200, Miguel said:
> > > 
> > > > We are migrating our mtas from ext3 to reiserfs and the
> > > > monitorization(a plugin) relies on the output of df -i, this
> > > > output is very diferent between those filesystems, because
> > > > reiserfs always return -1 as a total number of inodes.
> > > 
> > > The only reason you care about df -i is because on ext[23] and
> > > similar, it is possible to run out of inodes before you run out of
> > > files.  Since reiserfs allocates inodes as needed, you can't run
> > > out of them unless you run out of space - in which case your
> > > monitoring system should be flagging that.
> > 
> > No, we check this value to know how many files are in that
> > partition (well an aproximation). I hope that this is true for
> > ext2/3, isn't?
> > 
> 
> Would this help you?
> 
> # cat /proc/fs/reiserfs/md1/oidmap
> 
> At the end of oidmap is a line like this:
> total:  46 [46/972] used: 176147 [exact]
> 
> I think the used: 176147 is pretty close to what you want.  The number
> goes up and down when I create or remove files.  When I ran 'find /
> -xdev | wc -l' I got 174425.  I'm not sure where the extra 1,000+
> comes from.  
> 

i'm tring that on a 2.6.6 kernel and those files show bad info.

cat /proc/fs/reiserfs/hda7/oidmap
.
.
.
total:  166 [166/972] used: 764910 [exact]

bash-2.05b$ sudo find /disk/ | wc -l
128171


This method is usable on 2.6.x kernels?

proc info from this partition is attached.

tnx

> You would need to have CONFIG_REISERFS_PROC_INFO set in your kernel,
> and of course you would need to replace md1 with whatever device you
> were using.
> -- 
> Jonathan Briggs
> jbriggs@esoft.com
> 


-- 
La resistencia es fútil todos seréis asimilados


[-- Attachment #2: reiserfsproc --]
[-- Type: text/plain, Size: 8036 bytes --]

------------------- /proc/fs/reiserfs/hda7/bitmap ---------------------
free_block: 0
  scan_bitmap:          wait          bmap         retry        stolen  journal_hintjournal_nohint
              0              0              0              0              0              0              0
------------------- /proc/fs/reiserfs/hda7/journal ---------------------
jp_journal_1st_block: 	18
jp_journal_dev: 	hda7[0]
jp_journal_size: 	8192
jp_journal_trans_max: 	1024
jp_journal_magic: 	761207985
jp_journal_max_batch: 	900
jp_journal_max_commit_age: 	30
jp_journal_max_trans_age: 	0
j_1st_reserved_block: 	18
j_state: 	0
j_trans_id: 	1153730
j_mount_id: 	1245
j_start: 	2130
j_len: 	0
j_len_alloc: 	0
j_wcount: 	0
j_bcount: 	0
j_first_unflushed_offset: 	2127
j_last_flush_trans_id: 	1153728
j_trans_start_time: 	0
j_list_bitmap_index: 	2
j_must_wait: 	0
j_next_full_flush: 	0
j_next_async_flush: 	0
j_cnode_used: 	1
j_cnode_free: 	16383

in_journal: 	           0
in_journal_bitmap: 	           0
in_journal_reusable: 	           0
lock_journal: 	           4
lock_journal_wait: 	           0
journal_begin: 	           2
journal_relock_writers: 	           0
journal_relock_wcount: 	           0
mark_dirty: 	           1
mark_dirty_already: 	           0
mark_dirty_notjournal: 	           0
restore_prepared: 	           0
prepare: 	           1
prepare_retry: 	           0
------------------- /proc/fs/reiserfs/hda7/oidmap ---------------------
used: [ 1 .. b1abf )
free: [ b1abf .. b1ac0 )
used: [ b1ac0 .. b1be0 )
free: [ b1be0 .. b1be1 )
used: [ b1be1 .. b3d7e )
free: [ b3d7e .. b3d7f )
used: [ b3d7f .. b613f )
free: [ b613f .. b6140 )
used: [ b6140 .. b659b )
free: [ b659b .. b659c )
used: [ b659c .. ba846 )
free: [ ba846 .. ba847 )
used: [ ba847 .. ba871 )
free: [ ba871 .. ba872 )
used: [ ba872 .. baabe )
free: [ baabe .. baabf )
used: [ baabf .. baac0 )
free: [ baac0 .. baac2 )
used: [ baac2 .. baae4 )
free: [ baae4 .. baae5 )
used: [ baae5 .. baae6 )
free: [ baae6 .. baaf6 )
used: [ baaf6 .. baaf7 )
free: [ baaf7 .. baaf9 )
used: [ baaf9 .. bab0b )
free: [ bab0b .. bab0c )
used: [ bab0c .. bab1c )
free: [ bab1c .. bab1d )
used: [ bab1d .. bab1f )
free: [ bab1f .. bab22 )
used: [ bab22 .. bab23 )
free: [ bab23 .. bab24 )
used: [ bab24 .. bab25 )
free: [ bab25 .. bab27 )
used: [ bab27 .. bab43 )
free: [ bab43 .. bab44 )
used: [ bab44 .. bab4c )
free: [ bab4c .. bab4d )
used: [ bab4d .. bab82 )
free: [ bab82 .. bab83 )
used: [ bab83 .. bab84 )
free: [ bab84 .. bab87 )
used: [ bab87 .. bab88 )
free: [ bab88 .. bab8c )
used: [ bab8c .. bab8d )
free: [ bab8d .. baba0 )
used: [ baba0 .. baba1 )
free: [ baba1 .. baba2 )
used: [ baba2 .. baba3 )
free: [ baba3 .. babbb )
used: [ babbb .. babbc )
free: [ babbc .. bac1a )
used: [ bac1a .. bac1b )
free: [ bac1b .. bac1c )
used: [ bac1c .. bac1d )
free: [ bac1d .. bac21 )
used: [ bac21 .. bac22 )
free: [ bac22 .. bac23 )
used: [ bac23 .. bac24 )
free: [ bac24 .. bac25 )
used: [ bac25 .. bac26 )
free: [ bac26 .. bac27 )
used: [ bac27 .. bac28 )
free: [ bac28 .. bac2b )
used: [ bac2b .. bac2d )
free: [ bac2d .. bac2e )
used: [ bac2e .. bac32 )
free: [ bac32 .. bac34 )
used: [ bac34 .. bac35 )
free: [ bac35 .. bac36 )
used: [ bac36 .. bac38 )
free: [ bac38 .. bac39 )
used: [ bac39 .. bac71 )
free: [ bac71 .. bac7c )
used: [ bac7c .. bac7d )
free: [ bac7d .. bac7e )
used: [ bac7e .. bac7f )
free: [ bac7f .. bac80 )
used: [ bac80 .. bac81 )
free: [ bac81 .. bac88 )
used: [ bac88 .. bac8b )
free: [ bac8b .. bacde )
used: [ bacde .. bacdf )
free: [ bacdf .. bace7 )
used: [ bace7 .. bace8 )
free: [ bace8 .. bace9 )
used: [ bace9 .. bacea )
free: [ bacea .. baceb )
used: [ baceb .. bacec )
free: [ bacec .. baced )
used: [ baced .. bacee )
free: [ bacee .. bacf0 )
used: [ bacf0 .. bacf5 )
free: [ bacf5 .. bacf6 )
used: [ bacf6 .. bacf7 )
free: [ bacf7 .. bacfc )
used: [ bacfc .. bad01 )
free: [ bad01 .. bad03 )
used: [ bad03 .. bad04 )
free: [ bad04 .. bad05 )
used: [ bad05 .. bad06 )
free: [ bad06 .. bad07 )
used: [ bad07 .. bad09 )
free: [ bad09 .. bad0a )
used: [ bad0a .. bad0b )
free: [ bad0b .. bad0c )
used: [ bad0c .. bad0d )
free: [ bad0d .. bad0e )
used: [ bad0e .. bad0f )
free: [ bad0f .. bad10 )
used: [ bad10 .. bad11 )
free: [ bad11 .. bad12 )
used: [ bad12 .. bad13 )
free: [ bad13 .. bad14 )
used: [ bad14 .. bad15 )
free: [ bad15 .. bad16 )
used: [ bad16 .. bad17 )
free: [ bad17 .. bad18 )
used: [ bad18 .. bad19 )
free: [ bad19 .. bad1c )
used: [ bad1c .. bad1d )
free: [ bad1d .. bad1e )
used: [ bad1e .. bad21 )
free: [ bad21 .. bad22 )
used: [ bad22 .. bad23 )
free: [ bad23 .. bad24 )
used: [ bad24 .. bad25 )
free: [ bad25 .. bad26 )
used: [ bad26 .. bad27 )
free: [ bad27 .. bad28 )
used: [ bad28 .. bad29 )
free: [ bad29 .. bad2a )
used: [ bad2a .. bad2b )
free: [ bad2b .. bad2c )
used: [ bad2c .. bad2d )
free: [ bad2d .. bad2e )
used: [ bad2e .. bad2f )
free: [ bad2f .. bad30 )
used: [ bad30 .. bad31 )
free: [ bad31 .. bad32 )
used: [ bad32 .. bad33 )
free: [ bad33 .. bad36 )
used: [ bad36 .. bad39 )
free: [ bad39 .. bad3a )
used: [ bad3a .. bad3b )
free: [ bad3b .. bad3c )
used: [ bad3c .. bad3d )
free: [ bad3d .. bad3e )
used: [ bad3e .. bad3f )
free: [ bad3f .. bad40 )
used: [ bad40 .. bad41 )
free: [ bad41 .. bad42 )
used: [ bad42 .. bad43 )
free: [ bad43 .. bad44 )
used: [ bad44 .. bad45 )
free: [ bad45 .. bad46 )
used: [ bad46 .. bad47 )
free: [ bad47 .. bad4b )
used: [ bad4b .. bad56 )
free: [ bad56 .. bad85 )
used: [ bad85 .. bad86 )
free: [ bad86 .. bad89 )
used: [ bad89 .. bad8a )
free: [ bad8a .. bad8f )
used: [ bad8f .. bad90 )
free: [ bad90 .. ffffffff )
total: 	166 [166/972] used: 764910 [exact]
------------------- /proc/fs/reiserfs/hda7/on-disk-super ---------------------
block_count: 	2929846
free_blocks: 	212216
root_block: 	2052402
blocksize: 	4096
oid_maxsize: 	972
oid_cursize: 	166
umount_state: 	2
magic: 	 ReIsEr2Fs
fs_state: 	0
hash: 	r5
tree_height: 	5
bmap_nr: 	90
version: 	2
flags: 	1[attrs_cleared]
reserved_for_journal: 	0
------------------- /proc/fs/reiserfs/hda7/per-level ---------------------
level	     balances [sbk:  reads   fs_changed   restarted]   free space        items   can_remove         lnum         rnum       lbytes       rbytes     get_neig get_neig_res  need_l_neig  need_r_neig
0	            0            2            0            0         4424           21            0            0            0            0            0            0            0            0            0
1	            0            2            0            0          640          312            0            0            0            0            0            0            0            0            0
2	            0            2            0            0         3784          181            0            0            0            0            0            0            0            0            0
3	            0            2            0            0         8080            2            0            0            0            0            0            0            0            0            0
4	            0            0            0            0            0            0            0            0            0            0            0            0            0            0            0
------------------- /proc/fs/reiserfs/hda7/super ---------------------
state: 	REISERFS_VALID_FS
mount options: 	BORDER SMALL_TAILS 
gen. counter: 	0
s_kmallocs: 	0
s_disk_reads: 	0
s_disk_writes: 	0
s_fix_nodes: 	0
s_do_balance: 	0
s_unneeded_left_neighbor: 	0
s_good_search_by_key_reada: 	0
s_bmaps: 	0
s_bmaps_without_search: 	0
s_direct2indirect: 	0
s_indirect2direct: 	0

max_hash_collisions: 	0
breads: 	0
bread_misses: 	0
search_by_key: 	2
search_by_key_fs_changed: 	0
search_by_key_restarted: 	0
insert_item_restarted: 	0
paste_into_item_restarted: 	0
cut_from_item_restarted: 	0
delete_solid_item_restarted: 	0
delete_item_restarted: 	0
leaked_oid: 	0
leaves_removable: 	0
------------------- /proc/fs/reiserfs/hda7/version ---------------------
3.6 format	with checks off

Re: "df -i" support

[Dr. Giovanni A. Orlando]

Miguel wrote:

Dear Genius,

    Please adopt:

          df -hi

Thanks,
Giovanni

>Hi
>We have several machines running as MTA's using reiser for the mail
>queue, we usually check the number of files (well an approximation)
>using the output of "df -i" command on our monitorization plugins this
>may be not very acurate but is very fast and didn't affect the MTA's
>work.
>
>With reiser the output of df -i is no longer usable, we changed this to
>"find | wc -l" variations but it hurts the performance.
>
>There's a command to count the number of leafs (that's the name?) of a
>tree from a reiser partition? or other orientative value?.(we mount
>with notail)
>
>Tnx
>
>  
>


-- 

-- 

--
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

--

Re: "df -i" support

[Miguel]

On Thu, 13 May 2004 09:54:29 +0000
"Dr. Giovanni A. Orlando" <gorlando@futuretg.com> wrote:

> Miguel wrote:
> 
> Dear Genius,
> 
>     Please adopt:
> 
>           df -hi
from df(1):

....
-h, --human-readable
              print sizes in human readable format (e.g., 1K 234M 2G)
....

how will this help to our monitorization system? it's written in perl
and it's not a human.
read my e-mail.

> 
> Thanks,
> Giovanni
> 
> >Hi
> >We have several machines running as MTA's using reiser for the mail
> >queue, we usually check the number of files (well an approximation)
> >using the output of "df -i" command on our monitorization plugins
> >this may be not very acurate but is very fast and didn't affect the
> >MTA's work.
> >
> >With reiser the output of df -i is no longer usable, we changed this
> >to"find | wc -l" variations but it hurts the performance.
> >
> >There's a command to count the number of leafs (that's the name?) of
> >a tree from a reiser partition? or other orientative value?.(we mount
> >with notail)
> >
> >Tnx
> >
> >  
> >
> 
> 
> -- 
> 
> -- 
> 
> --
> 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
> 
> --
> 


-- 
La resistencia es fútil todos seréis asimilados