Strange delay in syscall close() on a large ext3-filesystem

[evilninja at gmx.net]

Volker Dose wrote:
> Good Morning,
> 
> If I do that strace cat <FILE> on another machine, I also see a close(1):

ok. i don't know if this is important at all, i just noticed it.

> 
> 
> --- snip ---
> write(2, "\n", 1
> )                       = 1
> close(1)                                = 0
> exit_group(1)                           = ?
> --- snip ---
> 
> And on another cyrus-mailserver close(3) and close(1):
> 
> --- snip ---
> ) = 1723
> read(3, "", 4096)                       = 0
> close(3)                                = 0
> close(1)                                = 0
> exit_group(0)                           = ?
> --- snip ---
> 
> 
> What exactly do this numbers mean? Filedescriptor? Isn't 1,2 and 3 for
> STDIN,STDOUT and STDERR?
> 
> The tune2fs-binary is just what you get, when youo install an SLOX and do
> the usual updates:
> 
> --- snip ---
> slox:~ # rpm -qf /sbin/tune2fs
> e2fsprogs-1.28-30
> --- snip ---
> 
> On my SLES-9 cyrus-mailserver:
> mail:/var/spool/imap/user/volker # rpm -qf /sbin/tune2fs
> e2fsprogs-1.36-6.2
> --- snip ---
> 
> 
> The kernen-version on SLOX is:
> --- snip ---
> slox:~ # uname -a
> Linux slox 2.4.21-295-smp #1 SMP Wed Jun 29 16:45:30 UTC 2005 i686 unknown
> --- snip ---
> 
> SLOX is based on United Linux 1, as you can see, a quite old distribution. I
> think it ios very close to SLES 8.

i really don't know what could've caused the long delays in close(). i'd
compile e2fsprogs-1.39-WIP-1210.tar.gz just to be sure no errors on the
fs are missed.

please respond on-list, so that the e2fs-gurus can comment....

Christian.
-- 
BOFH excuse #25:

Decreasing electron flux