ufs-discuss@opensolaris.org
[Top] [All Lists]

Re: [ufs-discuss] Re: File system full?

Subject: Re: [ufs-discuss] Re: File system full?
From: "Frank Batschulat (Home)"
Date: Wed, 21 Feb 2007 14:59:43 +0100
On Wed, 21 Feb 2007 14:31:32 +0100, Dennis <corovita@xxxxxxxxx> wrote:

ufs
magic   11954   format  dynamic time    Wed Feb 21 14:28:15 2007
sblkno  16      cblkno  24      iblkno  32      dblkno  760
sbsize  2048    cgsize  8192    cgoffset 64     cgmask  0xffffffc0
ncg     93      size    4530330 blocks  4461120
bsize   8192    shift   13      mask    0xffffe000
fsize   1024    shift   10      mask    0xfffffc00
frag    8       shift   3       fsbtodb 1
minfree 1%      maxbpg  2048    optim   time
maxcontig 7     rotdelay 0ms    rps     60
csaddr  760     cssize  2048    shift   9       mask    0xfffffe00
ntrak   48      nsect   128     spc     6144    ncyl    1475
cpg     16      bpg     6144    fpg     49152   ipg     5824
nindir  2048    inopb   64      nspf    2
nbfree  0       ndir    15704   nifree  381641  nffree  135498
cgrotor 52      fmod    0       ronly   0       logbno  1568
rolled  2       si      1       flags   0
version 2
fs_reclaim FS_RECLAIM
Dateisystem-Status gÃltig,
fsclean ist -3
BlÃcke verfÃgbar in jeder Drehposition

so you formerly claimed:

<snip>
unknon ufs NOTICE: alloc /: file system full
Thats impossile: There must be at least 90 GB free space
<snip end>

thats not possible since / is about 34.56 GB in size, since
you have fs->fs_size = 4530330 (number of blocks in fs)
* block size of 8k fs->fs_bsize = 8192 makes up a ~ 34.56 GB file system.

now the fstyp output does show that you do not have
an full sized file system blocks available - nbfree = 0,
the rest of the free space in this file system is made out of
fragments - nffree  135498.

the filesystem is indeed full from ufs perspective as there are
no free blocks available anymore, you will not be able to create a single file
that is larger  then "nbfree" * "bsize" (8k blocksize) because there aren't
enough free blocks. the reason for that is presumably lot of activity
on it in the past and the filesystem got fragmented enough or you
really do have some large file laying around sitting on the blocks.

You can use fsck(1M) instead to inspect the level of fragmentation:
At the very end, it reports (example):

35834 files, 467 used, 8691007 free (119 frags, 1086361 blocks, 0.0% fragmentation)

In order, those are the number of:
- inodes in use
- full-sized blocks in use
- fragments that are free
- fragments that are free but can not be part of a block due to
  fragmentation (aka isolated)
- full-size blocks that are free
- isolated fragments / total number of non-meta-data fragments

Note that fragments free should always equal isolated fragments plus
(fs_frag times full-size free blocks).  fs_frag is the number of fragments
in a block, you get this again from fstyp(1M) -v and field "frag".

du -hd / will list you files and the size they use up
and you may want to check this list of large consumers you're not aware of.


---
frankB
_______________________________________________
ufs-discuss mailing list
ufs-discuss@xxxxxxxxxxxxxxx

<Prev in Thread] Current Thread [Next in Thread>