xfs
[Top] [All Lists]

Re: Linux 2.4.22 XFS 1.3.1 reservation ran out

To: Jan Derfinak <ja@xxxxxxxxxxxx>
Subject: Re: Linux 2.4.22 XFS 1.3.1 reservation ran out
From: Eric Sandeen <sandeen@xxxxxxx>
Date: 01 Dec 2003 10:32:30 -0600
Cc: Nathan Scott <nathans@xxxxxxx>, linux-xfs@xxxxxxxxxxx
In-reply-to: <Pine.LNX.4.58.0311291838150.1157@xxxxxxxxxxxxxxxxxx>
Organization: Eric Conspiracy Secret Labs
References: <Pine.LNX.4.44.0310270844480.9654-100000@xxxxxxxxxxxxxxxxxxxxxx> <3FC6E5C4.2080009@xxxxxxxxxxxxxx> <20031129085715.A1872948@xxxxxxxxxxxxxxxxxxxxxxxx> <Pine.LNX.4.58.0311291838150.1157@xxxxxxxxxxxxxxxxxx>
Sender: linux-xfs-bounce@xxxxxxxxxxx
On Sat, 2003-11-29 at 17:41, Jan Derfinak wrote:

> With ikeep bonnie worked without error. Interesting is that before test df
> showed: 
> /dev/loop0              519488       144    519344   1% /mnt/mnt2
> 
> After test:
> /dev/loop0              519488    131340    388148  26% /mnt/mnt2
> 
> And there wasn't any file on the disk. Was the space consumed by inode
> table?

Yep, inodes are dynamically allocated as needed, so you're seeing them
take up space.  Without ikeep, there was some new code to remove unused
inode clusters.  (Nathan has since swizzled the mount options a bit, so
now keeping the clusters is default, and "noikeep" will test the removal
code).

Note that the above situation (disk space used for inodes) is really no
different than, say, ext3 - except that with ext3, you allocate all
those inodes (and use the space) at mkfs time, not runtime.

-Eric

-- 
Eric Sandeen      [C]XFS for Linux   http://oss.sgi.com/projects/xfs
sandeen@xxxxxxx   SGI, Inc.          651-683-3102


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