lkml.org 
[lkml]   [2000]   [Sep]   [21]   [last100]   RSS Feed
Views: [wrap][no wrap]   [headers]  [forward] 
 
Messages in this thread
/
Date
SubjectRe: fsck
From
   Date: 	Mon, 18 Sep 2000 12:06:22 +0200
From: octave klaba <oles@ovh.net>

I make the tests of fsck with raid-soft 2x18Go raid-1.
We crash the server to see how much time does fsck take
(power down) :)

- with standard /etc/rc.d/rc.sysinit of redhat 6.2
it takes 1h30 (with lot of inodes)

- i modified the options of fsck and erased -a.
In place I put -y
The fsck takes 8 minutes with log of inodes (more inodes ;) ).

My guess is that you just got lucky with your second fsck run. You
didn't actually use a controlled filesystem image for your tests, did
you? And if you just did the test without rebooting or flushing the
cache, you may have been working with a warm cache (i.e., disk blocks
already in the cache, so the system didn't have to read them in from the
system).

- what is the difference between -a et -y ?

For e2fsck, -a is the same as -p. The processing for -p and -y are the
same; the only difference comes when certain filesystem corruptions are
determined. With -p, certain filesystems corruptions that might require
human discretion to decide what the right way to fix things will cause
e2fsck to halt the filesystem check, whereas with -y e2fsck will blindly
try to fix problems in the most straightforward way.

- what is it so long with -a ?

My guess is because your test was flawed. You didn't give me enough
details for me to be sure how your test was flawed, but if I had to
guess, it's because you didn't flush the disk buffers before doing the
time tests. The standard way that I do my e2fsck time trials are:

e2fsck -Fftt /dev/XXXX

-F causes the disk buffers to be flushed before starting the time
trials.

-f forces a filesystem check

-tt causes e2fsck to print timing information on a pass-by-pass basis.

The other thing that you need to check is that if your filesystem has
corruptions/problems, that subsequent trials start with the filesystem
in the same state. (i.e., save a copy of the filesystme using dd, and
then do trials on the copy of the filesystem image, so that subsequent
trials can be done by using dd to make a fresh copy of the filesystem,
so that each successive e2fsck timing test have to do exactly the same
amount of work).

- Ted
-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.kernel.org
Please read the FAQ at http://www.tux.org/lkml/

\
 
 \ /
  Last update: 2005-03-22 12:38    [W:0.045 / U:2.068 seconds]
©2003-2020 Jasper Spaans|hosted at Digital Ocean and TransIP|Read the blog|Advertise on this site