Fw: Summary: df -k : discrepancy in the % displayed

And here is some light on the subject thrown by Anthony A. D. Talltree.
Thanks Anthony for filling up the gaps,

Ramesh
=================================================
A couple of gaps in the info that others sent:

o The primary reasoning for this space reservation is that on the
traditional BSD filesystem, write of new or growing files really slow
down as the filesystem approaches being full and the filesystem code has
to look harder for free space.  The 10% is historical -- it was probably
decided twenty years ago by observing the behavior of the hardware of
that time, with the load and usage patterns of that time.  Hardware has
improved since then, and various tweaks have been made to the UFS
filesystem, hence Sun's adaptive value.

o It is entirely reasonable to set the minfree to 0 on filesystems that
are either
a) substantially read-only (say, a jumpstart tree)
 or
b) preallocated, eg. modern news spools are big monolithic pre-allocated
files.  In past years, I would also do this for filesystems that held
SunOS 4.x diskless swap files.  In both cases, blocks are constantly
written *within* the preallocated files, but there's no allocation of
new blocks within the filesystem.

Since 36G SCSI disks can be had new for $180 and used for $60, there
isn't much incentive to do it for routine read/write filesystems.
=====================================================

----- Original Message -----
From: "Ramesh Pathak" <ramesh.pathak@vinciti.com>
To: <sunmanagers@sunmanagers.org>
Sent: Thursday, August 28, 2003 5:03 AM
Subject: Summary: df -k : discrepancy in the % displayed


> Let me thank all the great guys who responded to my query so fast.
> Here are the excerpts from the mails that I received.
> The reason is "minfree" taking some bytes out from the user.
> And Paul also suggests that fragmentation may add to it further.
>
> Thanks to all of you,
> Ramesh
>
> Michael Schulte
> ===========
> The standard file system reserves about 10% of the kbytes for efficiency;
it
>    is unavailable to anyone except root.  Notice that the used+avail !=
> kbytes.
>
> VAleriy
> ========
> If you created your filesystem with default options, 10% of the space
> was reserved for root only access. So, in reality, you have to divide it
> by (0.9*963869), i.e.
> 963869kb available total, take 10% for root only writes and you'll get
> 867462.1, now  185680/867462.1 = 21%
> see man newfs section for -m option
>
> Tom Payerle
> =========
> On Unix filesystems, there is typically some portion of a filesystem which
> is reserved for root only; i.e. once a filesystem is filled to a certain
> point,
> only root can write to it.  This is useful for stuff like / and /usr to
> enable
> things to work pretty well even when filesystems get "full".  I think the
> mkfs/newfs commands default to about 5% for this root reservation.  Some
> systems (I forget whether Solaris has this) have tunefs type commands that
> can alter it once the filesystem is up and running, others it must be done
> at creation time.
>
> In your root case,
> used (185680) + avail (720357) = 906037 or 57832 less than in kbytes
field.
> (which is 6% of the total).
>
> Note that 185680/906037 = 20.5%
>
> This 5% or so reservation is probably not worth bothering with on a 1-2GB
> system disk or partitions (/, /usr, /var, etc).  It probably is worth
messing
> with when dealing with 10 GB or more data only partitions (e.g. the system
> can function pretty well if partition was missing), although to be honest
> even in those cases I usually haven't done anything about it.
> Tom Payerle
> Dept of Physics payerle@physics.umd.edu
> University of Maryland (301) 405-6973
> College Park, MD 20742-4111 Fax: (301) 314-9525
> ID16314 Explanation of file system usage as reported by df and fsck
>
> Paul Richards
> ==============
> :df output
> :   Filesystem            kbytes    used   avail capacity ...
> :/dev/sd2d            1961966 1745945   19825    99%   ...
>
> :1961966 kbytesFile system size
> :1745945kbytes usedAmount used
> :19825 kbytes freeEqual to (size less minfree%) less used
> :99%Equal to used as % of (size less minfree%)
>
> so minfree can affect the answer
> also fragmentation can have an effect
>
> :For example, in a file system that creates predominantly 5k files on an
> :8k block size there will be many 3k frags free and never used. In the
> :extreme case this would result in a file system that is effectively
> :full despite only 5/8 of the filesystem being used.
>
> Jay Lessert
> ========
> minfree.
>
> Looks like / was built with 6% minfree and /usr with 2% minfree.
>
> /: 185680/(0.94 * 963869) = 20.5%
> /: 2263070/(0.98 * 3099093) = 74.51%
>
> You can see current minfree (and lots of other interesting information
> about a file system with:
>
> % sudo fstyp -v /dev/rdsk/c0t0d0s0
>
>
> Crist J. Clark
> ========
>
> Read df(1),
>
>      -k    Print the allocation in kbytes. The output consists of
>            one  line  of information for each specified file sys-
>            tem. This information includes the file  system  name,
>            the  total  space  allocated  in  the file system, the
>            amount of space allocated to existing files, the total
>            amount  of  space  available  for  the creation of new
>            files by unpriviledged users, and  the  percentage  of
>                     ^^^^^^^^^^^^^
>            normally  available  space that is currently allocated
>            to all files on the file system. This option overrides
>            the -b, -e, -n, and -t options.
>
>
> And newfs(1M),
>
>            -m free
>                  The minimum percentage of free space to maintain
>                  in   the   file  system  (between  1%  and  99%,
>                  inclusively). This space is off-limits to normal
>                  users.  Once  the  file system is filled to this
>                  threshold,  only  the  super-user  can  continue
>                  writing  to  the file system. This parameter can
>                  be subsequently  changed  using  the  tunefs(1M)
>                  command.
>
>                  The default is  ((64  Mbytes/partition  size)  *
>                  100),  rounded  down  to the nearest integer and
>                  limited between 1% and 10%, inclusively.
>
>
>
> Lars
> ====
> man newfs(1m), -m option.
>
>  capacity = used / ( used + avail )
>
>  used + avail < kbytes.
>
>
> --
> Omnibiblious, adj.:
>         Indifferent to type of drink.  "Oh, you can get me anything.
> I'm omnibiblious."
>
> Eugene
> =========
> Hi
>
> By default a certain amount of space is allocated as "reserved for root
use
> only". This space is also commonly referred to as "slosh space".
>
> This can be modified with tunefs (see man page).
>
> The calculation takes this into account. Typically this is between 1 and
10
> %, depending on the FS size.
>
>
> -----Original Message-----
> The % available is based on the 90% level of capacity:
>
> /: ( 185680 / (.90 * 963869)) = 21%
>
>
> Mark
> ==========
> thats because when you do a defualt newfs it save 10% for superuser.. man
> newfs and look for the -m option:
>
>           -m free        The minimum percentage of free space  to
>                          maintain in the file system.  This space
>                          is off-limits to normal users.  Once the
>                          file system is filled to this threshold,
>                          only the super-user can continue writing
>                          to  the file system.  This parameter can
>                          be  subsequently   changed   using   the
>                          tunefs(1M) command.  The default is 10%.
>
> So with a newfs -m 3 /dev/dsk/c0t0d0s0 will give you 7% more disk space
> for that partition...
>
> Hope this helps...
>
> Kevin
> =====
>
>      The capacity percentage reported by "df -k" is the amount of user
> writable space that is in use.  What you're calculating is the total
> amount of space that is in use.  By default, UFS filesystems reserve 10%
> of the space in the filesystem that can only be written to by root (as an
> aside, this can and should be changed when large filesystems are created;
> 10% of a 1 TB filesystem is a lot of wasted space).  This is also why you
> can see filesystems that are more than 100% full.  HTH...
>
> ==================================
>
> Kevin Buterbaugh - Systems Engineer
> LifeWay - www.lifeway.com
> Man tunefs
> Look at the -m Option and it was clearer ;-)
>
>
> --
> Michael Schneider                               MM-ITconsulting
> Sun Certified Network Administrator    Ph: +49(0)700
>
>
>
>
>
>
> Ramesh C Pathak
> Vinciti Networks Pvt. Ltd.,
> #1109, 24th Main Road,
> JP Nagar 1st Phase,
> Bangalore - 560 078
>
> Ph: +91-80-6556830
> Fax: +91-80-6556820
>
> email: ramesh.pathak@vinciti.com
> URL: http://www.vinciti.com
>
> This information is intended only for the use of the individual to whom it
> is addressed and may contain information that is privileged, confidential
> and exempt from disclosure under law. If you are not the intended
recipient,
> you are hereby notified that any use, dissemination, distribution or
copying
> of this communication is strictly prohibited.
> _______________________________________________
> sunmanagers mailing list
> sunmanagers@sunmanagers.org
> http://www.sunmanagers.org/mailman/listinfo/sunmanagers
_______________________________________________
sunmanagers mailing list
sunmanagers@sunmanagers.org
http://www.sunmanagers.org/mailman/listinfo/sunmanagers