RE: Fewer partitions are better (Re: Disk Layout/Partitioning Practices)
by Leam Hall
> David;
>
> I like the logic of why you would not put all data in /. I'm about to
rebuild a system and wonder how you would recommend breaking / up?
>
> Paul
>
> Sorry about the top post ... Working with Outlook :(
>
> -----Original Message-----
> From: dballester(a)kernpharma.com [mailto:dballester@kernpharma.com]
> Sent: Thursday, January 29, 2004 1:25 AM
> Hi:
> I understand your point of view but I disagree with you. I'm
sysadmin and I NEED to have system data and user data in three or more
physical partitions ( at least /boot, / and /where_user-app_data_is ).
The reason is disponibility. I'm totaly agree with LVM and RAM reasons
that you exposed but having all data in / is dangerous.
> Think about a damaged filesystem. In parititioned systems, if
the damaged filesystem is user data or /boot, I can unmount it easily,
repair it, and mount it again. If all system except /boot are in /, I
need to shutdown the machine, startup in 'recovery mode', repair and
start machine.
> In this machine, for example we can have, internal dns, dhcp
server, cups ( printing ) and samba server. Using different partitions,
samba user data can be unmounted, and printig, dns and dhcp will not be
affected. With only / I can have all people in company stopped for a
while.
> <snip>
I prefer to break up user space for safety reasons. Some useres don't
know they have too much data on the machine until they choke it. If your
user stuff is on / then they can prevent normal system operations.
Normally I use /, /boot, /usr, /var, and /export; with the latter
getting home directories, shared filesystems, etc. /usr and /var/*could*
go on /, but I'm mostly used to keeping them seperate.
ciao!
leam
--
20 years, 3 months
Kernel build problem
by mjwestkamper
Whilst building a customized kernel, during make install unresolved symbols
are reported. A google shows this has been around for a while, however I
could not find a resolution. Any suggestions?
I will gladly provide the .config or any other information that will help.
Mike
======================
tools/build -b bbootsect bsetup compressed/bvmlinux.out CURRENT > bzImage
Root device is (3, 3)
Boot sector 512 bytes.
Setup is 4926 bytes.
System is 934 kB
sh -x ./install.sh 2.4.22-1.2149.nptl30012004 bzImage
/usr/src/linux-2.4.22-1.2149.nptl/System.map ""
+ '[' -x /root/bin/installkernel ']'
+ '[' -x /sbin/installkernel ']'
+ exec /sbin/installkernel 2.4.22-1.2149.nptl30012004 bzImage
/usr/src/linux-2.4.22-1.2149.nptl/System.map ''
depmod: *** Unresolved symbols in
/lib/modules/2.4.22-1.2149.nptl30012004/kernel/drivers/char/drm/sis.o
depmod: sis_malloc_Ra3329ed5
depmod: sis_free_Rced25333
grubby fatal error: unable to find a suitable template
make[1]: Leaving directory
`/usr/src/linux-2.4.22-1.2149.nptl/arch/i386/boot'
[root@auxfs linux-2.4]#
=======================
20 years, 3 months
Fewer partitions are better (Re: Disk Layout/Partitioning Practices)
by Keith Lofstrom
Ron Herardian <rherardi(a)gssnet.com> writes:
> When installing everything and allowing for future updates and packages
> I am using the following disk layout:
... 7 partitions listed ...
It may be time to rethink multiple partitions. For legacy reasons,
I build my systems with multiple partitions, but if I had it to do
over again I would probably do it with 3 partitions:
/boot (because it has to be small)
swap (about 2 GB)
/ (everything else, including multiple drives if LVM)
For the average system, the swap probably should be smaller, but I
run huge, poorly designed CAD apps for days that tend to fill VM.
With RAM so damned cheap these days, I agree with other folks that
make swap a lot smaller than 2x physical RAM. Paging more than a
few hundred megabytes of swap is just too damned slow, regardless
of the amount of RAM that you happen to have in your system.
If I had a problem with users writing huge files that filled the disk,
I would put quotas on the individual users, not on their partition as
a whole. If anything goes wild, and fills up /, the system is in
trouble anyway, and repair time is not significantly improved by
limiting the fillup to one of many partitions. In fact, a lot of
problems happen because /tmp fills, or /var/mail fills and /var/log
doesn't work. I subscribe to the Andrew Carnegie maxim "Put all
your eggs in one basket - then WATCH THAT BASKET".
My past excuses for multiple partitions were: (1) limited disk sizes
and (2) managing backup tapes. Both are invalid now. With large
drives and LVM, there is no practical limit on partition sizes for
most systems. With disk-to-disk random access backup, there is no
need for complicated partition arrangements to fit data onto small,
slow tapes. I control backup frequencies and depth by directory,
not by partition.
On mild excuse for retaining multiple partitions is to minimize
boot time - if you are running a journalling file system, it will
occasionally delay booting for significant time to fsck one or two
of the partitions. But if you have a very large /home, the time
saved by peeling off a bunch of smaller partitions is not that
significant. There are probably better ways to schedule fsck.
I expect strong opinions differing from the above; no doubt I've
forgotten something. Newbies should keep following this thread and
see where I've erred. But a lot of the reasons we do things are
traditions that stem from past restrictions that no longer apply.
We should acknowledge these changes in our system designs.
Keith
--
Keith Lofstrom keithl(a)ieee.org Voice (503)-520-1993
KLIC --- Keith Lofstrom Integrated Circuits --- "Your Ideas in Silicon"
Design Contracting in Bipolar and CMOS - Analog, Digital, and Scan ICs
20 years, 3 months
good video player for fedora
by mitch epling
is there one ? . i tried downloading mplayer and the package i download
when i go to install it it goes and then says it needs the package i am
installing to proceed but its the one i am installing its asking for
20 years, 3 months
idle HD spin down standby/sleep, laptop mode?? (seagate 7200.7)
by Kevin Bowen
I'm trying to get fedora to spin down my primary drive when idle, but I
can't seem to get it to go into standby mode. When I use hdparm to put it
into 'sleep' mode, it does sleep very briefly, waking up after a few
seconds. When I use standby mode (either forced or with an -S idle time
auto-standby) it doesn't seem to ever enter standby at all. I'm not exactly
clear on how the new "laptop mode" works into this, but I have tried it both
with and without this mode enabled (including running the laptop-mode.sh
script) with no apparent change in behavior.
My drive is a serial ATA seagate 7200.7, running on a SATA raid card (not
using any raid features) - might this be the source of the problem?
Any ideas appreciated
kevin(a)ucsd.edu
20 years, 3 months
IPTABLES doesn't work
by smoothmilk
Why doesn't redhat-config-securitylevel's iptables rules work?
If I turn off EVERYTHING (www, ftp, ssh, etc) and save, and even
manually restart iptables (# /sbin/service iptables restart) other
computers on my network can access www (even on weird, non-standard
ports with http servers on them) ftp, ssh, etc.
So whats the point of even including that tool if it doesn't do
anything? I dont understand how it just flat out doesn't work. I have no
idea how iptables works, and because there's no documentation out there
for beginners who just want a script that's for eth0 with a simple www,
ssh and ftp server(s), Im stuck using rh's tools, which don't do
anything. there's no security here.
Can anyone help?
20 years, 3 months
3C905-TX problems
by Michael Bernstein
Hi,
I'm having problems with my first Linux installation (so expect some
naivite). The Ethernet card - a 3C905-TX on an MSI K7N2G mobo - is a real
mare!! Let me say that all works well under WinXP. The NIC is connected to
the Cable modem (ftom ntl in the UK).
I had loads of trouble getting the board to go "active" at all, but
finally managed after I turned off kudzu ("chkconfig kudzu off" and a
reboot). Now I can ping and browse, but everything goes at a snail's pace.
If I ping the BBC website I get 50% packet loss.
I've spent a LOT of time with this now so help would be greatly
appreciated.
Many thx,
Mike
PS I tried to work with the on-board nVIDIA ethernet but that wasn't
recognised, and nVIDIA don't supply a driver for core 2.4.22-1.2115
_________________________________________________________________
Express yourself with cool new emoticons http://www.msn.co.uk/specials/myemo
20 years, 3 months
Xine kills X
by Jonathan Villa
I just installed Xine via yum and it kills X once it's started...
the rpms I have installed are
xine-lib-1.0.0-0.5.rc3.fr
xine-0.9.23-1.fr
xine-lib-devel-1.0.0-0.5.rc3.fr
all I can capture via xine 2>errors.txt is
XIO: fatal IO error 104 (Connection reset by peer) on X server ":0.0"^M
after 2296 requests (2295 known processed) with 31 events
remaining.^M
** ERROR **: Must shutdown ORB from main thread
aborting...
any ideas?
20 years, 3 months
D-Link DWL-650+ and acx100 on FC1
by Dexter Ang
Can anyone confirm that the hardware "D-Link DWL-650+" works with FC1?
This is a WiFi 802.11b CardBus PCMCIA. As I've read through some page,
it works with the acx100 drivers. I understand that this kernel module
is precompiled in dag.wieers' repository. Is there a generic SRPM of
this, as I'd rather have the option of (re)compiling myself, so I
don't have to wait for the precompiled rpms when there is a new kernel
out.
Thanks
dex
20 years, 3 months