Basic Administration Concepts

• Welcome to linuxzoo
  • System Administration
    • Basic Administration Concepts

Disks and Partitions

Disks

• /dev/hda - primary disk on first IDE controller
• /dev/hdb - slave on first IDE controller
• /dev/hdc - primary disk on second IDE controller
• /dev/hdd - slave on second IDE controller
• /dev/sda - lowest numbered SCSI device
• /dev/sdb - next lowest SCSI device
• ...

Partitions

• Rather than use the whole disk for one purpose...
• Split disk up into chunks.
• The chunks are known as partitions.
• Partitions can be primary or secondary.
• This is partially a hang-over from when DOS could only handle 4 partitions...

> sfdisk -l /dev/sda

Disk /dev/sda: 8924 cylinders, 255 heads, 63 sectors/track
Units = cylinders of 8225280 bytes, blocks of 1024 bytes, counting from 0

   Device Boot Start     End   #cyls    #blocks   Id  System
/dev/sda1   *      0+     12      13-    104391   83  Linux
/dev/sda2         13    1287    1275   10241437+  83  Linux
/dev/sda3       1288    1542     255    2048287+  82  Linux swap
/dev/sda4       1543    8923    7381   59287882+   f  W95 Ext'd (LBA)
/dev/sda5       1543+   1797     255-   2048256   82  Linux swap
/dev/sda6       1798+   8923    7126-  57239563+  83  Linux

$ cat /etc/fstab

• When the system boots the fstab file tells the kernel what filesystems to load

  /dev/ubd/0    /    		ext3  	defaults        1 1
  none          /dev/pts    	devpts gid=5,mode=620  0 0
  none       	 /dev/shm    tmpfs  defaults        0 0
  none          /proc       	proc  	defaults        0 0
  none          /sys        sysfs  	defaults        0 0
  /dev/ubd/1    swap     	swap  	defaults        0 0
  

$ df

Filesystem           1K-blocks      Used        Available Use% Mounted on
/dev/sda2             10080520     3142968   6425484    33%	/
/dev/sda1             101086         9665         86202        11% 	/boot
none                     1038660       0              1038660     0% 	/dev/shm
/dev/sda6             56340828      3853984  49624868   8% 	/home

• df -h is also useful, translating bytes in MB or GB as appropriate

*UML

• In UML, there are no IDE or SCSI drives.
• The disks are called /dev/ubd/n where n is a number
• They are actually implemented by files in the host operating system, but this is hidden from you.
  • /dev/ubd/0 is /
  • /dev/ubd/1 is swap

Disk Usage

• If you want to find out how much disk space a directory is using, the "du" command does this easily.

  $ du -s /usr/lib
  477464  /usr/lib
  $ du -sh /usr/lib
  467M    /usr/lib
  

• "-s" is useful, otherwise it tells you about all subdirectories too.
• "-h" puts it into human readable form.

Linux Boot Process

Booting to kernel

• From switch-on:
  • PC BIOS selects a boot disk
  • BIOS loads the boot block and executes it.
  • This loads a stage 1 boot loader.
  • Stage 1 loads stage 2 loader.
  • Linux loader (e.g. Grub, lilo) runs
  • Operator selects from loader menu
  • Kernel loaded with device ramdisk

$ cat /etc/grub.conf

default=1
timeout=10
splashimage=(hd0,0)/grub/splash.xpm.gz
title Fedora Core (2.6.6-1.435.2.3)
        root (hd0,0)
        kernel /vmlinuz-2.6.6-1.435.2.3 ro root=LABEL=/ rhgb quiet
        initrd /initrd-2.6.6-1.435.2.3.img
title Fedora Core (2.6.5-1.358)
        root (hd0,0)
        kernel /vmlinuz-2.6.5-1.358 ro root=LABEL=/ rhgb quiet
        initrd /initrd-2.6.5-1.358.img

Startup Commands

• As linux boots, it runs various system scripts.
• Eventually it runs one for your standard "runlevel".
• The runlevel startup enables the services (like ssh and apache") which you may want to start.
• Such startup scripts all live in: /etc/init.d/
• For example, apache is looked after in:

  /etc/init.d/httpd
  

/etc/init.d/*

• The scripts in init.d can:
  • Start
  • Stop
  • Restart
  • Reload
  • + a few others

Run levels

• The run level determines what init.d files run.
• As you enter a run level services not running which should run at that run level start.
• As you leave a run level services which should not be running at the new run level stop.
• What start and stop are determined by the soft links found in the /etc/rd?.d directories.
• Usually all we need to know is the stardard runlevel is 5.

$ ls /etc/rc5.d

K01yum           K35vncserver  K74ypxfrd            S13portmap    S80sendmail
K05saslauthd   K35winbind     K89netplugd         S14nfslock      S90crond
K10dc_server   K45named      K95kudzu             S18rpcgssd     S90xfs
K10psacct        K50netdump   K96init.cssd          S19rpcsvcgssd S95anacron
K12dc_client    K50snmpd      S00microcode_ctl  S20random     S95atd
K12mysqld      K50snmptrapd  S04readahead_early  S25netfs   S96init.cssd
K20nfs            K50tux           S06cpuspeed         S26apmd        S96readahead
K24irda           K54dovecot    S08iptables           S28autofs       S97messagebus
K25squid         K70aep1000   S09isdn                S44acpid         S97rhnsd
K34dhcrelay     K70bcm5820  S10network          S55sshd          S99local
K34yppasswdd  K74ntpd        S12syslog             S56rawdevices S99mdmonitor
K35dhcpd         K74ypserv     S13irqbalance       S56xinetd

S/K priority service-name

• S99mdmonitor :

• $ ls -l /etc/rc5.d/S99mdmonitor
  lrwxrwxrwx  1 root root 19 Jul 27 13:00 S99mdmonitor -> ../init.d/mdmonitor
  

• Starts at priority 99 - runs last
• Start mdmonitor at this runlevel

Link management

• We used to have to create the soft links ourselves to manage the run levels. This is disgusting!
• In Redhat, chkconfig does this job.

  > chkconfig -list mdmonitor
  mdmonitor  0:off 1:off 2:on 3:on 4:on 5:on 6:off
  > chkconfig --levels 345 mdmonitor off
  > chkconfig -list mdmonitor
  mdmonitor  0:off 1:off 2:on 3:off 4:off 5:off 6:off
  

The syslog

• Clicking on the syslog console output link on the control window finishes with:

   Starting system message bus: [ OK ]
   Starting  mdmonitor : [ OK ] 
  

• That confirms mdmonitor was started
• You can also read this from within linux using the "dmesg" command.

The xinetd super-daemon

• Some of the services (e.g. sshd) are processes.
• They start running from an rc script.
• They wait on their own for comms.
• They terminate only when the machine does down.
• Some people say this wastes resources.
• The super-server concept was born.

XINETD

• Xinetd waits for requests from the internet.
• From the requests it works out what program would like to deal with that request.
• It then starts that program running and gives it the waiting requests.
• In this way resources are only used if someone actually requests access to a particular service.
• The approach of xinetd is not as popular as services started explicitly from init.d.
• The virtual machines use xinetd to control telnet.
• If you connect to your VM with telnet, xinetd starts up the telnet daemon for you.
• Services from xinetd usually start "in.".
• Telnet is "in.telnetd"
• /etc/xinetd.d/ contains all the services it manages.

$ cat /etc/xinetd.d/telnet

service telnet
{
        flags           	= REUSE
        socket_type 	= stream
        wait            	= no
        user            	= root
        server          	= /usr/sbin/in.telnetd
        log_on_failure  += USERID
        disable         = no
}

Terminating a process

• If you know the process id (the PID) of a program you can terminate it quickly and easily
• You send it a message (a signal) to tell it to end.
• The message to end now is called SIGKILL.

  > ps aux | grep sshd
  root 1796 ...… /usr/sbin/sshd
  > kill -s SIGKILL 1796
  

User Management

User Management

• A wide topic...
  • Adding/Removing/Changing current users
  • Default Scripts
  • Global Scripts

Manual Creation

• User entries in passwd,shadow, group,gshadow.
• Home directory in /home.
• Copy basic .files into their home directory.
• Make new user own their own directory and files.

$ adduser gordon

• This does all the magic for you.
• It copies the default .files from /etc/skel/

  > ls -a /etc/skel/
  .bash_logout  .bash_profile  .bashrc  .gtkrc  .kde
  

• In bash, .bashrc is executed in non-login shell, and .bash_profile in a login shell.

Skel files

• These files are the basic .files created for a new user.
• Users are free to edit these when they log in.
• This allows them to control their own path, env, and other settings (such as aliases).
• However, if you install a new package which needs something set for each user at login, editing all these copies by hand would be tiresome.

$ ls /etc/profile.d

colorls.csh  gnome-ssh-askpass.csh  	krb5.csh  less.csh  vim.csh
colorls.sh    gnome-ssh-askpass.sh	krb5.sh    less.sh    vim.sh
glib2.csh     kde.csh                          	lang.csh  qt.csh      which-2.sh
glib2.sh       kde.sh                         	lang.sh    qt.sh

• If you log in with bash, all the .sh files are executed before your .files
• If you log in with csh, all the .csh files are executed before your .files.

$ cat /etc/profile.d/vim.sh

if [ -n "$BASH_VERSION" -o -n "$KSH_VERSION" -o -n "$ZSH_VERSION" ]; then
  # for bash, pdksh and zsh, only if no alias is already set
  alias vi >/dev/null 2>&1 || alias vi=vim
fi

• I.e. if this is bash, and you have not set an alias for "vi", then set one to run "vim" when you type "vi".

Example

• Create a user jim, in group staff
• But how to set the group?
• You could do:

  $ man adduser
  

• Usually commands also take the flag "-h"

  $ adduser -h
  adduser: invalid option -- h
  Usage: useradd [options] LOGIN
  
  Options:
    -b, --base-dir BASE_DIR       base directory for the new user account
    ...
    -g, --gid GROUP               force use GROUP for the new user account
    ...
  
  
  $ adduser jim -g staff
  $ tail -1 /etc/passwd
  jim:x:502:100::/home/jim:/bin/bash
  $ grep 100 /etc/group
  staff:x:100:
  

Moving a uid or gid

Useful Commands

$ chown jim.staff filename
$ chown jim filename
$ chgrp staff filename

When a User logs in

• When a user logs in the appropriate . files are executed (.login, .cshrc, etc).
• If you want to change to a different user, you could log out and log in again, or you could do
  • su - gordon (change to the gordon user)
  • su - (change to root)
• Without the "-", you still change users, but the . scripts don't get executed.
• To go back to the previous user, press CTRL-D

FILE SEARCHING

A file CONTAINING something

• You are looking for a file containing "gordon"
• You think it is in /etc/ something

  $ grep "gordon" /etc/*
  /etc/group:gordon:x:500:
  /etc/group-:gordon:x:500:
  /etc/gshadow:gordon:!!::
  /etc/gshadow-:gordon:!!::
  /etc/passwd:gordon:x:500:100:Dr Gordon ...
  

A FILENAME containing something

• You know somewhere in /etc there is a filename with the word "host" in it.

  $ find /etc -name ‘*host*'
  /etc/hosts.deny
  /etc/ghostscript
  /etc/ssh/ssh_host_dsa_key.pub
  

• Find can find on a range of things, not just names. Other things include sizes, permissions, types, ownership, and combinations of tests.

Find to do something

• Usually find prints the things which match.
• You can get it to execute instead.
• Here you want to find all files called core, and delete them:

  $ find . -name core -print -exec rm {} \; 
  

• I didn't invent the syntax, so don't blame me...

Discussion

• A user keeps getting logged out each time they log in... why?

Discussion

• A user find their ls command is broken... why?

Discussion

• Here are some past exam questions you should now be able to answer:

Question 1

• What is the function of "su", and what is the difference between "su - gordon" and "su gordon"?

Question 2

• What type of files would you expect to find in /sbin?

Question 3

• Consider the following line:

  gordon:x:44:
  

  In which file in /etc would you expect to see such a line, and what does it mean?

Question 4

$ mkdir temp
$ cd temp/
$ mkdir txt.txt/
$ cd txt.txt/
$ touch hello
$ cd ..
$ ls *.*