Transcript Administração de Sistemas

Administração de Sistemas
(ASIST)
TP 1 (English version)
LINUX servers installation
1
LINUX Distributions
The LINUX operating system kernel implements the most basic and important
functions like device management, process management and memory
management, it interacts with the HARDWARE directly.
You can say “the kernel is the operating system”. The LINUX kernel is in
constant development, its free and open source.
Above all the kernel provides a stable platform for upper level software to work
free from low level details.
SOFTWARE
In order to take advantage of the kernel capabilities many
LINUX Kernel
other auxiliary programs are required to obtain a fully
HARDWARE
functional system.
Some of this auxiliary programs are critical for the system start-up, for example
they provide means to load the kernel in memory and give it the control over
the system. Some other provide ways to install a fresh operating system.
The set KERNEL + PROGRAMS is known as “Distribution”, distributions are
not all free as they may include commercial software.
2
LINUX distributions installation
Most current distributions provide high level user friendly
assisted installation programs. They protect the user from
questions to which they don't know the correct answers.
The assisted installation solution, almost options free, has the advantage of bringing
many common users to LINUX however for administrators has some disadvantages.
The decisions installation programs take without asking may undermine some special
propose the system may have and special requirements needed.
Fortunately in most cases there is an expert mode where less decisions automatic and more
questions are asked.
On the other hand almost everything can be reconfigured later after the initial installation.
However the division of the disk in partitions is critical because it won’t be easy to
change later, after the initial installation.
3
Disks and partitions in LINUX
In LINUX (and UNIX in general) most HARDWARE resources are internally
identified by objects in the “/dev” system folder. Disks are identified in different
forms according to the kind of disk interface they use.
IDE: /dev/hda /dev/hdb /dev/hdc /dev/hdd
SCSI or SATA : /dev/sda /dev/sdb /dev/sdc /dev/sdd /dev/sde
…
Most disks (not CD/DVD) are divided in partitions. Partitions are logical divisions
of the disk. They are defined in a special zone in the start of the disk (MBR) called
“partitions table”.
Each partition in a disk is absolutely independent of others in the same disk, each
may contain totally different kind of data with different formats, it may even
happen that different partitions in the same disk are used by different operating
systems. For each partition there is a identifier of the format in the partition table.
UNIX identifies partitions in the disk by appending a number starting with “1” for
the first partition. Example: /dev/hdc2 = 2ª partition of the master disk on the
second IDE controller.
4
Partition tables – shown by “fdisk” (LINUX)
HOST1# fdisk /dev/sda
Command (m for help): p
Disk /dev/sda: 73.4 GB, 73407868928 bytes
255 heads, 63 sectors/track, 8924 cylinders
Units = cylinders of 16065 * 512 = 8225280 bytes
Device
Boot
Start
End
Blocks
/dev/sda1 *
1
101
811251
/dev/sda2
102
8924
70870747+
/dev/sda5
102
202
811251
/dev/sda6
203
711
4088511
/dev/sda7
712
4757
32499463+
/dev/sda8
4758
8797
32451268+
/dev/sda9
8798
8924
1020096
Id
83
5
82
83
83
83
83
System
Linux
Extended
Linux swap
Linux
Linux
Linux
Linux
HOST12# fdisk /dev/hda
Command (m for help): p
Disk /dev/hda: 20.0 GB, 20003880960 bytes
255 heads, 63 sectors/track, 2432 cylinders
Units = cylinders of 16065 * 512 = 8225280 bytes
Device
Boot
Start
End
Blocks
/dev/hda1 *
1
13
104391
/dev/hda2
14
778
6144862+
/dev/hda3
779
1415
5116702+
/dev/hda4
1416
2432
8169052+
/dev/hda5
1416
1925
4096543+
/dev/hda6
1926
1990
522081
/dev/hda7
1991
2432
3550333+
Id
83
83
83
5
83
82
83
System
Linux
Linux
Linux
Extended
Linux
Linux swap
Linux
5
Partition table – shown in Windows
A partition is a zone of the disk defined in the partition table by a starting point and
a finish point, all in between belongs to the partition.
For each partition there is a identifier in the partition table which identifies the file
system format used in the partition.
6
Partitions formatting
The main advantage of creating partitions in a disk is that each of them can be used
without any concern about other partitions usage.
A partition cam be used by a operating system to store data, most often organized
in folders and files. The way data is organized and stored in the partition is known
as file system format.
There are several file system formats. Currently Microsoft Windows uses mostly
NTFS, but the old FAT can also be used. In Linux the current standard file system
formats are EXT2 and EXT3.
The file system stores several information about the objects it holds, namely the
location of their data in the partition and their properties like name and ACL. All
this information about objects is stored on structures in the partition itself. The
process of creating those structures is known as formatting.
After formatting a partition it will appear to be completely empty and ready to store
new data.
7
File system access
The way operating systems provide access to file systems resident in different
portions of different disks varies, in DOS/Windows systems drive letters are used
with this purpose. Letters “A:” and “B:” are normally reserved for floppy disks, so
the first disk file system typically gets the “C:” letter.
In Linux (and Unix in general) another principle is used, one file system (partition)
is chosen to be the root file system. Other file systems of any kind can latter be
integrated on the root file system by a operation called mounting.
Mounting creates logical associations controlled by the operating system between
empty folders and file systems stored in other disks and partitions.
After the mounting any
request to the originally
empty folder will be
redirected by the operating
system.
/etc
/ (root)
/dev/sda1
/mnt
/usr
/dev/hdb
/mnt/cdrom1
Mounting
/usr/local
/dev/hda2
8
Partitions and file systems planning
Automatic installation programs try to find the best partition and file system
configuration, to achieve this they perform a small analysis o the disk size and
features.
In most cases the decisions taken will be the correct ones, however the installation
program lacks the knowledge about the exact final purpose of the system.
Due to this lack of knowledge some decisions might reveal later to be wrong.
Fortunately in most cases an
opportunity is given to change the
method in order for the user to be able
to do things his way.
In the sample by directly managing the
partition table. In other cases by
selection on of several partitions
model.
9
Partitions size
Because later will be hard to change the partitions sizes, the best would be to have
them with the correct size in the initial installation.
We can state that the best efficiency in disk usage is achieved by using a single
partition for the entire disk. That's true because partitions create static limits which
he system cant overcame, situations may rise in which one partition is full and other
partitions on the same disk are almost empty .
However on the safety point of view this limits might be a good thing, they provide
a way to keep different parts of the system independent on disk storage. If on part
of the system (sub-system) fills it partition than wont disturb other sub-systems.
For instance, if user areas (homes) are placed in a separate partition then if a
users fills his work area, that wont affect the system because it doesn’t
depend on that partition.
(but it would affect other users because the partition is shared between them )
10
Partition size and quotas
In order to make the disk storage of each user independent form other users
creating a partition for each user is not reasonable.
The way to achieve that purpose is the quota system, when active the kernel will
keep a constant control on the space each user or group is taking in the partition.
Quotas can be defined at partition level only
Because quotas work at partition level, the way partitions are created will affect how many
different quota subsystems may exist.
Several quota subsystems may be required, for instance:
- User area quota (home)
- WEB area quota
- E-MAIL quota
- System partition quota
For each a different partition will be required.
Other point against homes in the system partition is that when a user’s quota is exhausted he
wont be able to make any use of the system.
11
Root partition sizing
The purpose of the root partition is to provide
a base for the kernel startup, it must contain all
the necessary files for the initial boot. Only
latter it will be possible o mount other file
systems.
The root partition must have enough space to keep all the base operating system,
including objects that will most likely grow during the system's lifetime, for instance
log files, configuration files and installed software.
Due to the current disk sizes its senseless to use root partitions smaller than 8
gigabytes, in the future smaller sizes could become a problem.
Increasing the available disk space its simple, disk are cheap and one can always be
added. The administrator may then create new partitions, format and finally mount
them on the existing root file system. Finally some cosmetic may be required in
order for folders appear to be on the right place, symbolic links will solve that.
12
Swap partition sizing
The SWAP partition is used to implement
virtual memory. Because disks are very slow
when compared with real memory, virtual
memory should be seen as a last resource
solution that avoids “out of memory” errors.
In other words, when talking about servers, the central memory (RAM) should be
such that virtual memory won't be used very often. Given that principle there is no
point in having very big swap capacity. As a reference value its usual to create a swap
partition with a size equal to the system’s RAM..
Because Linux can manage several swap partitions at the same time, as far as there is
free space in a disk it’s always possible to increase the swap space by adding new swap
partitions.
13
File system type
The Linux operating system supports a wide
variety of file systems, currently the most used is
ext3 (third extended file-system).
Most Linux installation programs of current
distributions defaults to ext3.
As for disk partitions, distribution installation programs provide a way to interact and
select other kind of file system to be used. Some interesting alternatives to ext3 are
ReiserFS and XFS.
All this modern file systems are journaling file systems, they log changes made to the
disk, in case of system crash the system wont have to check the hole disk, only the
areas where changes where happening.
This means than after a system crash the time required to bring the system up again
will be very short when compared with non journaling file systems like ext2.
14
Initial root and modules
Even tow a file system is supported by the Linux kernel than doesn’t mean it can be
used as initial root.
Because Linux supports a very wide range of features, their support cant be all
include in the kernel as it would grow to big.
Some kernel features require kernel modules kept in separate files (kernel loadable
module - KLM).
The loading of a module requires a working file system to hold the module file, this
means the initial root file system must be fully supported by the kernel, the support
cant be provided by a module. One way to overcome this limitation is the use of a
initial ram disk as temporary root file system, the required modules can be placed in
the ram disk. Later the root file system can be changed to a real disk partition.
Current distributions kernels fully support ext2 and ext3 without modules, however the same
is not necessarily true for other files systems like ReiserFS and xfs.
15
Formatting details
The file system formatting operations uses specific programs, for example with the
base name “mkfs” followed by the file system type, like “mkfs.ext3” or “mkfs.ext2”.
Automatic
Linux
installation
programs call this external programs
with the arguments tuned up for the
partition characteristics, namely its
size and kind of use it will have.
mke2fs [ -c | -l filename ] [ -b block-size ] [ -f fragment-size
] [ -g blocks-per-group ] [ -i bytes-per-inode ] [ -j ] [ -J
journal-options ] [ -N number-of-inodes ] [ -n ] [ -m
reserved-blocks-percentage ] [ -o creator-os ] [ -O
feature[,...] ] [ -q ] [ -r fs-revision-level ] [ -E extendedoptions ] [ -v ] [ -F ] [ -L volume-label ] [ -M last-mounteddirectory ] [ -S ] [ -T filesystem-type ] [ -V ] device [ blockscount ]
Among the arguments some may affect the performance, for instance higher values
for bock-size and bytes-per-inode may increase performance for some applications.
In the sample above the “-T” option allows a generic specification about the size of
the files it is intended to store. The command will then automatically define the best
values for other important arguments.
16
Network interface configuration
In a server the best solution for interface configuration is static and manually defined
Dynamic network configuration, typically by using the DHCP protocol, creates a
external dependency for the system. If the DHCP server is not working the our
server wont work either. If possible all dependencies like this should be avoided in a
server system.
Manual static network configuration requires cooperation between the server
administrator and the network administrator, the later will have to provide the
required information:
-
IP address + network mask + default gateway
-
Local DNS domain name and local name servers addresses
The major disadvantage of static manual configuration is the need to manually
change the server configuration each time the network administrator makes a
change on that information.
17
Install modes for LINUX
Linux distributions provide several ways to perform the
installation. The installation process is itself based on a
Linux kernel.
The typical way to start the installation is to boot the machine from a bootable
CD/DVD. Because the installation system is Linux based, any technique than can be
used to start Linux can be used to start the installation process as well. For instance a
directly for DOS/WINDOWS or using a boot EPROM.
Because distributions want to as complete as possible, nowadays they have become
quite large with up to a 20 CD’s, however most of the software include wont be
installed.
The so called “network installation” requires just a CD or even smaller support
media, after starting the installation system it uses the network (connected to the
internet) to get the requested from public storage places created to this purpose.
18
Linux installation - hardware
The wide variety and constant development of the hardware is
the responsible for many of the problems in the installation
process.
Very recent hardware may not be fully supported by Linux, even by he latest
distributions.
Anyway the correct procedure should be to chose the hardware for the operating
system and not the way around. Most Linux distributions provide listings of
compatible and tested hardware.
Another available option is the acquisition of a box ready to work, hardware and
preinstalled operating system.
Disk controllers may present a problem, if they aren't supported by the installation
system then the installation becomes impossible.
19