Increasing inodes without recreating file system

Damian Menscher menscher at uiuc.edu
Wed Jul 27 15:59:29 UTC 2005


On Wed, 27 Jul 2005, Robin Laing wrote:
> Robin Bowes wrote:
>> nodata wrote:
>>> 
>>> I don't think you can do it online.
>> 
>> That was my understanding too - I thought I'd ask here as a last resort :(

Random idea (assuming your vpopmail is less than 5 gig or so):

Try shrinking your /usr partition, and create a new partition there with 
the -T news option.  Then move /home/vpopmail over to that partition. 
That will free up nearly half of your inodes in /home, which may be 
sufficient for your needs.

BTW, you make for a great example.  I've always (for some convoluted 
definition of "always") known it was possible to run out of inodes, but 
never saw anyone do it.

> As I am planning on doing the same thing in the very near future with a RAID 
> array, this is of interest to me.  I will have many files that are in the 10k 
> to 200K range, graphics and ogg/mp3's/flac files.
>
> Does the default number of inodes work with this or should I run mkfs.ext3 
> with the -i option and specify a small number?
>
> Or be save and use the -T  news option and be safe?

I think the defaults will work for you, as they appear to create one 
inode for each 8k on the disk.  Since your smallest files will be 10k, 
it will probably work, but possibly be slightly tight.  Giving the -T 
news option would be safer, of course, just in case you end up using it 
differently than your predictions.  For example, what if you decided to 
add a small text file to describe each image/song?  That would require 
double the inodes.  Advance planning is critical here, since you can't 
change your mind later.

> Is there a maximum number of inodes allowed per filesystem?

I think the issue is fsck time.  When you fsck, it needs to scan every 
inode.  So for really huge filesystesm that could take an insanely long 
time (think of hours of downtime).  I confronted this issue about a year 
ago when setting up a 3.5TB filesystem.  After waiting about an hour for 
a 1.8TB filesystem to fsck, I'd learned my lesson that having 233M 
inodes is not necessarily a good thing.  Especially given that only 1% 
were being used.  So I tweaked the parameters, and now have only 28M 
inodes for the 3.5TB filesystem.  This was based on expected usage of 
the filesystem, and then multiplying by 4 for safety.  It looks like my 
estimates were close -- the filesystem is currently using 66% of the 
space and 13% of the inodes.  And it will probably fsck within 15 
minutes.

Warning: don't say fsck time doesn't matter unless you're wearing 
asbestos underwear!

Damian Menscher
-- 
-=#| Physics Grad Student & SysAdmin @ U Illinois Urbana-Champaign |#=-
-=#| 488 LLP, 1110 W. Green St, Urbana, IL 61801 Ofc:(217)333-0038 |#=-
-=#| 4602 Beckman, VMIL/MS, Imaging Technology Group:(217)244-3074 |#=-
-=#| <menscher at uiuc.edu> www.uiuc.edu/~menscher/ Fax:(217)333-9819 |#=-
-=#| The above opinions are not necessarily those of my employers. |#=-




More information about the fedora-list mailing list