Yo Tiger,

>From experience, its basically a roll of the dice if your file system gets
corrupted or not during a power failure, and the extent of the damage is
also another roll of the dice.  I've seen servers come up again without a
hiccup (especially those with journalled ones) and i've seen utter
devastation even with journalled fs (exact symptoms as you described but
even worse).

Most default journalling setups will only ensure that the integrity of the
file system (as far as its structure aka metadata goes) is intact. It
won't mean your data is actually _written_ on the disk.  I've tried and
seen reiserfs '/var/log' partitions get mounted normally, but when you
look at the actual log files... they contain garbage.  This is potentially
bad for stuff like databases, or .. <Cringe> MP3 files.

An option i believe for your setup would be to use the data=journal mode
in ext3, mount all your filesystems with the 'sync' option to ensure data
really gets written to your disk.  Oh, and make sure you turn off the
write cache of your hard disk also (hdparm -W0)

Contrary to the expctation that speed will actually suffer due to this...

http://archives.postgresql.org/pgsql-performance/2002-09/msg00013.php

Seems to say the opposite.

Also make sure you use the latest kernel prepatches because Andrew Morton
fixed a data coruption bug in 2.4.20 for ext3 + data=journal mode.

HTH




On Sat, 25 Jan 2003, Gerald Timothy Quimpo wrote:

> On Saturday 25 January 2003 01:26 pm, Lemuel Tomas wrote:
> > on redhat systems there is a file you can edit to put in fsck parameters,
>
> i just looked on my notebook (mandrake 98% of the time, windows only
> to play Age of Empires II :), and in /etc/sysconfig there is a file named
> autofsck.  there i can specify autofsck.
>
> > word of caution: some people dont like to autofsck becuase of the potential
> > data loss they MAY result, you've better have a backup protocol in place to
> > save your data.
>
> hehe, i don't understand fsck enough to give it intelligent answers anyway
> when i do a manual fsck.  i just say yes to anything.  doing auto-fsck will
> be more of the same except i don't need to be there :).  thanks a lot for
> the pointer!  that'll be very useful.
>
> generally only the filesystem with /var (its own filesystem or somewhere
> other than / because of this) is the one that needs to be manually fsck-ed.
> i don't think i've ever had to fsck any of the others, even / (which has
> /tmp in it).  the box is mostly just a dialup server which is sometimes
> used for surfing and games.  there's not a lot of data there, and the
> data that is, is on a filesystem that doesn't get touched a lot.  so i'm
> not worried about data loss.  the real data *does* get backed up to
> another computer hard drive.
>
> > hope this helps,
>
> oh, lots, i think.  thanks again.
>
> tiger
>
>

_
Philippine Linux Users Group. Web site and archives at http://plug.linux.org.ph
To leave: send "unsubscribe" in the body to [EMAIL PROTECTED]

Fully Searchable Archives With Friendly Web Interface at http://marc.free.net.ph

To subscribe to the Linux Newbies' List: send "subscribe" in the body to 
[EMAIL PROTECTED]

Reply via email to