 |
  OS Lab Linux |
Linux Resources & Hints
RedHat and SuSE distribution brought Linux real mainstream exposure in 1998 and 1999, since then Linux-distribution got the acknowledgement in the server-market as well on the desktop-world, yet there it still has to catch up, specially in regards of devices drivers, which is unfortunatly a matter of device developing companies to also support Linux.
I personally ran it for servers and workstation but since May 2000 only on some dedicated machines where Linux is required such as
otherwise for servers and workstation I prefere nowadays FreeBSD, even for workstation purposes, it is just more convienent: install, upgrade apps or kernel, performance and overall system-integrity.
- VMware,
- Win4Lin,
- UserModeLinux,
- and few other 'linux-only' approaches;
Something many people aren't aware, Linux isn't a complete Operating System as such, it is just the kernel. Every distribution developed its own UNIX-like disk-layout arround the Linux-kernel. In this light, Linux "suffers" from this diversity:
RedHat RedHat is probably best known from all distributions, but this actually doesn't mean anything about quality necessarly, it is average.
RedHat.Com RedHat's site
SuSE SuSE is very well known in Europe and provides the largest collection of compiled packages installable from the CD, which is very handy to save download time. Comperable with RedHat-distribution.
SuSE.Com SuSE's site
Debian Debian is a non-profit undertake with support of packages. The installation isn't that user-friendly (as of version 2.2, and even 3.0) as no hardware is recognized automatically as SuSE or RedHat does, yet the collection of packages make this distribution very valuable over the commercial distributions mentioned above.
Debian.Org Main site DebianPlanet.Org News site Few things to know after install:
- use dselect to explore the available packages, or apt-cache search term; if you run X11 install gnome-apt.
- if you are used to *bsd-like port disk-layout, download debports which gives you a nice /usr/ports/ directory-structure, make install, deinstall and clean is supported. Also consider to check linuxports@sourceforge.
- if you fail to install .deb files (due missing dh_testdir as example), do apt-get install debhelper.
- /etc/apt/sources.list has the sources (cdrom, ftp) of the packages, use apt-get update to get up-to-date package list.
- server/client: to share the .deb files export from the server /var/chache/apt/archives/ and also /etc/apt/ to all clients, then at least the packages are shared in common.
Other distributions There are other distributions like Slackware and Mandrake (derived from RedHat), each one with its own flavour and features. See the full collection of Linux-distributions.
Download
LinuxISO.Org ISO Images of common Linux distributions Tiny Linux Dists micro-dists
IMHO I think the Linux-distributions should unify and provide one or two main-stream distributions in order to become competetive in the entire OS-market. For now SuSE and RedHat are leading, but quality-wise they are far below any *BSD distribution (e.g. FreeBSD) in my eyes: there are no packages (rpm doesn't catch up), there is no comformity between the different dists (fall 2000):
LinuxBase.Org Standardizing of LINUX GoboLinux.Org Completely UNIX directory layout
Some fragments I found on the net and discovered by myself:
New Swap Space
% dd if=/dev/zero of=/your/swapfile bs=1024 count=swap_size % chmod 600 /your/swapfile % mkswap /your/swapfile swap_size % swapon /your/swapfile Maximum swap_size call man mkswap to read more
Also add an entry in /etc/fstab:
/your/swapfile none swap sw 0 0
Open Files Edit /usr/src/linux/include/linux/limits.h under Kernel-2.0
#define OPEN_MAX 1024 /* # open files a process may have */ and recompile kernel.
With Kernel 2.2 you can
echo 32768 > /proc/sys/fs/file-max echo 65536 > /proc/sys/fs/inode-max
More Files
echo 8192 >/proc/sys/kernel/file-max echo 32768 >/proc/sys/kernel/inode-max put it into /etc/rc.d/rc.local
Memory Server: Memory is cheap, upgrade it to the max, e.g. 128MB at least, more is better when you run a web-server so swapping is never required in the best case. Workstation: I use GIMP and PicArt daily, which means huge RGBA images are handled, for that reason I upgraded my workstation to 768MB to have it running smoothly on the Athlon 1GHz (was using PPro 200 with 256MB before).
Mac Disks To read mac disks (floppy or zip) under Kernel-2.2 you may acquire: compile it, and copy hfs.o into /lib/modules/2.0.xx/fs/ and then call insmod hfs and mount -t hfs /dev/sdb /mnt/zip (as example)
HFS support kernel module
Memory Filesystem (RAM Disk) To speed up disk-performance you might copy disk-content on a memory-based filesystem called RAMFS (Kernel-2.4) or RAMDISK (Kernel-2.2), under *BSD system it's called MFS. Kernel 2.4
You need to have RAMFS enabled in your kernel, and then:
% mkdir /mnt/mfs % mount -t ramfs ramfs /mnt/mfs With kernel 2.4 the memory-fs under Linux grows and shrinks as required like under FreeBSD
Kernel 2.2
You need to have RAMDISK enabled in your kernel, and then /dev/ram* should and then /dev/ram* should exist already, read man ram otherwise. To setup and enable a ramdisk:
% mke2fs /dev/ram0 % mkdir /mnt/mfs % mount /dev/ram0 /mnt/mfs
The size of ramdisks are define via the kernel, or can be adjusted in the lilo.conf:
: image=/boot/vmlinuz label=linux root=/dev/hda2 ramdisk_size=10000
Software RAID Software RAID is an easy way to connect two or more hard-disks together and improve read/write speed:
Linux 2.4: RAID How to setup software RAID
Linux has been suffering a real state-of-the-art filesystem as ext2 fulfills the basic needs, but it's a pain to recover once your server crashed (can happen very rarely), but when it happens it just takes forever to bring an ext2-filesystem back into consistancy.
Ext3 With ext3 you can change any ext2 to ext3 on-the-fly (without reformating the disk): More infos see at Using ext2 fs in 2.4 kernels and Red Hat: ext3 information.
- mounted disk: tune2fs -j /dev/hd... does the job, a .journal is created in the root directory of the mounted device
- unmounted disk: change /etc/fstab the ext2 to ext3, the journal is not seen.
Misc
ReiserFS ReiserFS JFS by IBM Journaling FS developed by IBM (opensource) XFS by SGI Another Journaling Filesystem (part of SGI's OSS-initiative Interview With the People Behind JFS, ReiserFS & XFS Article at OSnews Realtime Data Mirroring Linuxfocus article using IBM's FAM (file alteration monitor)
Have "IP Aliasing" in kernel enabled (and maybe need to recompile the kernel) then add in /etc/rc.d/rc.local:
echo "adding virtual homes" /sbin/ifconfig eth0:0 domain2.com /sbin/ifconfig eth0:1 domain3.com
You may have to add also:
/sbin/route add -host domain2.com dev eth0:0 /sbin/route add -host domain3.com dev eth0:1
Of course the DNS needs to be valid and resolveable so these additional virtual domains work, check dnstool.
A typical linux procedure is to download the most recent kernel and compile it.
Kernel.Org Archive of all kernels Extract it in /usr/src/ (usually creates linux/ itself) and then use one of the three choices:
% make config # terminal based config (not recommended) % make menuconfig # terminal based config (recommended, needs ncurses) % make xconfig # x11 based config (recommended) and walk through all options. Then call
% make dep % make # make kernel % make install # install kernel (check /boot/* and /etc/lilo.conf) % make modules # make modules % make modules_install # install modules
If you ever tried *BSD system, you realize how messy this Linux kernel/lilo thing truly is:
Once again we recommend Debian/Linux as it behaves best with a clean UNIX system-layout.
- deselect SYMTAB export (one of the first config questions in kernel config)
- pay close attention of your /vmlinuz which may point to /boot/vmlinuz or /boot/vmlinuz-x.x.x
- keep backup of your existing kernel (e.g. cp vmlinuz vmlinuz.old)
- edit /etc/lilo.conf, and run /sbin/lilo as root
UML (User Mode Linux) is truly a nice piece of software as it allows to run multiple Linux-kernels and filesystem (guest) within on a Linux-machine (host), a feature mainframes have:
UML (User Mode Linux) Running Linux in user-mode(space) of Linux; check also UML-user list archive
- Download a root-filesystem e.g. root_fs_debian2.2_small
and make a sym-link
ln -s root_fs_debian2.2_small root_fs
- Download the UML-kernel e.g. linux-2.4.0-test12
and execute the kernel
./linux-2.4.0-test12 login as root (password 'root' too).
The advantage is obvious:
and beside, it prooves the Linux design being consistant in itself. The file-system is a flat-file and easily be backed-up or moved arround on other machines.
- great test-bed for other Linux variants
- fastest virtualization (faster than vmware or plex86)
- advanced resource accounting: nice the guest-Linux appropriatly on the host
- true virtualized web-hosting
More Infos
freeVSD Virtual Server Deamon (VSD): not really virtualization but 'chroot'ed environments: ssh, httpd, sendmail - solution for ISPs S_Context: Virtual Servers Another approach S/390 & Linux Good article at LinuxWorld.Com Linux/390 S/390 mainframe running Linux Hercules S/390 Emulator S/370 - ESA/390 emulator for Linux; running Linux/390 on the emulator itself. LINE Running linux-binaries under WinXX FAUMachine Open Source Virtual Machine Ensim Commercial package Virtuozzo Linux in Linux, commercial product
Here the procedure step-by-step for a Debian/Linux-2.2 (for other distributions it may work, but the directory layout may look slightly different) install:Then unplug the 2nd disk (hdc) and put it into your other system and look Linux booting.
- assuming your 2nd disk is /dev/hdc, make two partitions (or more with different mounts) with fdisk /dev/hdc:
hdc1 280MB type 83 hdc2 42MB type 82 (swap) then call
mke2fs /dev/hdc1 mkswap /dev/hdc2
- then transfer Linux live-system to your 2nd disk:
mkdir /mnt/newdisk mount /dev/hdc1 /mnt/newdisk cd /mnt/newdisk cp -a /sbin . cp -a /bin . cp -a /lib . cp -a /etc . cp -a /var . cp -a /dev . cp -a /sbin . cp -a /usr . cp -a /root .
- and then the special dirs
mkdir proc tmp chmod 1777 tmp mkdir home
- and the last copying:
cp -a /boot . ln -s boot/vmlinuz vmlinuz
- as next LILO: edit /mnt/newdisk/etc/lilo.conf, again, my 2nd dist is /dev/hdc, these specs are important:
boot=/dev/hdc disk=/dev/hdc bios=0x80 # make hdc be hda, useful for our case root=/dev/hda1 # here again the hda referenced this way you can install lilo on /dev/hdc, but it will treat it as hda (see LILO-HowTo), call lilo then like this:
lilo -r /mnt/newdisk -C etc/lilo.conf
- then edit /mnt/newdisk/etc/fstab:
/dev/hda1 / ext2 defaults,errors=remount-ro 0 1 /dev/hda2 none swap sw 0 0 proc /proc proc defaults 0 0
- other files to edit:
- /mnt/newdisk/etc/hosts
- /mnt/newdisk/etc/hostname
- /mnt/newdisk/etc/network/interfaces
This "Live-System Install" is particularly useful to install Linux medias like PCMCIA ATA disk, or other medias which aren't supported using the CD-Install (see Stylistic-1000 Linux install).
FAI (Fully Automatic Installation) for Debian GNU/Linux Installing Linux on a cluster
NewsForge.Com Very good tech-news site SlashDot.Org News site on open-source stuff mainly Linux Mall Buy stuff you need Linux Gazzette Online news-magazine XFree86 Free X-server since Version 4.x becoming the standard XIG Reliable x-server only worth to buy when you have an odd graphic card
 OS Lab | FreeBSD |
All Rights Reserved - (C) 1997 - 2008 by The Labs.Com
Top of PageThe Labs.Com