solaris 9 student guide part 2 sa299 phần 4 pps

The AutoFS file system enables you to do the following: ● Mount file systems on demand ● Unmount file systems automatically ● Centralize the administration of AutoFS mounts through the u

Trang 1

Configuring AutoFS

Objectives

The AutoFS file system provides a mechanism for automatically mountingNFS file systems on demand and for automatically unmounting these filesystems after a predetermined period of inactivity The mount points arespecified using local or distributed automount maps

Upon completion of this module, you should be able to:

● Describe the fundamentals of the AutoFS file system

● Use automount maps

The following course map shows how this module fits into the currentinstructional goal

Figure 7-1 Course Map

Managing Swap Configuration

Managing Crash Dumps and Core Files

Configuring NFS ConfiguringAutoFS

Managing Virtual File Systems and Core Dumps

Trang 2

Introducing the Fundamentals of AutoFS

AutoFS is a file system mechanism that provides automatic mountingusing the NFS protocol AutoFS is a client-side service The AutoFS filesystem is initialized by the /etc/rc2.d/S74autofsautomount script,which runs automatically when a system is booted This script runs the

automountcommand, which reads the AutoFS configuration files andalso starts the automount daemonautomountd The automountddaemonruns continuously, mounting and unmounting remote directories on anas-needed basis

Whenever a user on a client computer running theautomountddaemontries to access a remote file or directory, the daemon mounts the remotefile system to which that file or directory belongs This remote file systemremains mounted for as long as it is needed If the remote file system isnot accessed for a defined period of time, theautomountddaemonautomatically unmounts the file system

The AutoFS service mounts and unmounts file systems as requiredwithout any user intervention The user does not need to use themount

and umountcommands and does not need to know the superuserpassword

The AutoFS file system enables you to do the following:

● Mount file systems on demand

● Unmount file systems automatically

● Centralize the administration of AutoFS mounts through the use of aname service, which can dramatically reduce administration

overhead time

● Create multiple mount resources for read/write or read-only filesystems

Trang 3

The automount facility contains three components:

● The AutoFS file system

● The automountddaemon

● The automountcommand

Figure 7-2 The AutoFS Features

AutoFS File System

An AutoFS file system’s mount points are defined in the automount maps

on the client system After the AutoFS mount points are set up, activityunder the mount points can trigger file systems to be mounted under themount points If the automount maps are configured, the AutoFS kernel

module monitors mount requests made on the client If a mount request ismade for an AutoFS resource not currently mounted, the AutoFS servicecalls the automountddaemon, which mounts the requested resource

RAM

Automount Maps

Master map Direct map Indirect map

Special map

automount -v AutoFS

automountd

Trang 4

The automountd Daemon

The/etc/rc2.d/S74autofsscript starts theautomountddaemon at boottime Theautomountddaemon mounts file systems on demand andunmounts idle mount points

Note – Theautomountddaemon is completely independent from the

automountcommand Because of this separation, you can add, delete, orchange map information without having to stop and start theautomountd

daemon process

The automount Command

Theautomountcommand, called at system startup time, reads the mastermap to create the initial set of AutoFS mounts These AutoFS mounts arenot automatically mounted at startup time, they are the points underwhich file systems are mounted on demand

Trang 5

Using Automount Maps

The file system resources for automatic mounting are defined inautomount maps Figure 7-3 shows maps defined in the /etcdirectory

Figure 7-3 Configuring AutoFS Mount Points

The AutoFS map types are:

● Master map – Lists the other maps used for establishing the AutoFSfile system Theautomountcommand reads this map at boot time

● Direct map – Lists the mount points as absolute path names This

map explicitly indicates the mount point on the client

● Indirect map – Lists the mount points as relative path names This

map uses a relative path to establish the mount point on the client

● Special – Provides access to NFS servers by using their host names

NFS Client

"venues"

/

auto_master

/net -hosts [options]

/home auto_home [options]

/- auto_direct [options]

auto_direct /opt/moreapps pluto: /export/opt/apps

auto_home

Ernie mars:/export/home/ernie Mary mars:/export/home/mary etc

Trang 6

The automount maps can be obtained from ASCII data files, NIS maps,NIS+ tables, or from an LDAP database Together, these maps describeinformation similar to the information specified in the /etc/vfstabfilefor remote file resources.

The source for automount maps is determined by theautomountentry inthe/etc/nsswitch.conffile For example, the entry:

automount: files

tells theautomountcommand that it should look in the/etcdirectory forits configuration information Usingnisinstead offilestellsautomount

to check the NIS maps for its configuration information

Configuring the Master Map

The auto_mastermap associates a directory, also called a mount point,with a map Theauto_mastermap is a master list specifying all the mapsthat the AutoFS service should check Names of direct and indirect mapslisted in this map refer to files in the /etcdirectory or to name servicedatabases

Associating a Mount Point With a Map

The following example shows an /etc/auto_masterfile

# cat /etc/auto_master

# Master map for automounter

#

+auto_master

/net -hosts -nosuid,nobrowse

/home auto_home -nobrowse

Trang 7

The general syntax for each entry in the auto_mastermap is:

mount point map name mount options

where:

Note – The plus (+) symbol at the beginning of the+auto_masterline inthis file directs the automountddaemon to look at the NIS, NIS+, orLDAP databases before it reads the rest of the map If this line iscommented out, only the local files are searched unless the

/etc/nsswitch.conffile specifies that NIS, NIS+, or LDAP should besearched

mount point The full path name of a directory If the directory

does not exist, the AutoFS service creates one, ifpossible

map name The name of a direct or indirect map These maps

provide mounting information A relative pathname in this field requires AutoFS to consult the

/etc/nsswitch.conf file for the location of themap

mount options The general options for the map The mount

options are similar to those used for standard NFSmounts However, thenobrowse option is anAutoFS-specific mount option

Trang 8

Identifying Mount Points for Special Maps

There are two mount points for special maps listed in the default

The two mount points for special maps are:

Note – The-xfnmap provides access to legacy FNS resources Supportfor FNS is scheduled to cease with this release of the Solaris OE

The-hosts map Provides access to all resources shared by NFS

servers The resources being shared by a server aremounted below the/net/hostname directory, or, ifonly the server’s IP address is known, below the

/net/IPaddress directory The server does nothave to be listed in the hosts database for thismechanism to work

The-xfn map Provides access to resources available through the

Federated Naming Service (FNS) Resourcesassociated with FNS mount below the/xfn

directory

Trang 9

Using the /net Directory

Shared resources associated with the hosts map entry are mounted belowthe /net/hostnamedirectory For example, a shared resource named

/documentationon hostsys42is mounted by the command:

# cd /net/sys42/documentation

Using the cdcommand to trigger the automounting ofsys42’s resourceeliminates the need to log in to the system Any user can mount theresource by executing the command to change to the directory thatcontains the shared resource The resource remains mounted until apredetermined time period of inactivity has occurred

Adding Direct Map Entries

The /-entry in the example master map defines a mount point for directmaps

#

+auto_master

/- auto_direct -ro

The /-mount point is a pointer that informs the automountfacility thatthe full path names are defined in the file specified by map_name(the

/etc/auto_directfile in this example)

Note – The/-entry is not an entry in the default master map This entryhas been added here as an example The other entries in this examplealready exist in the auto_masterfile

Even though themap_nameentry is specified asauto_direct, the

automountfacility automatically searches for all map-related files in the

/etcdirectory; therefore, based upon the automountentry in the

/etc/nsswitch.conffile, theauto_directfile is the

/etc/auto_directfile

Trang 10

Note – An NIS or NIS+ master map can have only one direct map entry A

master map that is a local file can have any number of entries

Creating a Direct Map

Direct maps specify the absolute path name of the mount point, thespecific options for this mount, and the shared resource to mount Forexample:

# cat /etc/auto_direct

# Superuser-created direct map for automounter

#

/apps/frame -ro,soft server1:/export/framemaker,v5.5.6

/opt/local -ro,soft server2:/export/unbundled

/usr/share/man -ro,soft server3,server4,server5:/usr/share/man

The syntax for direct maps is:

key [ mount-options] location

where:

key The full path name of the mount point for the direct

maps

mount-options The specific options for a given entry

location The location of the file resource specified in

server:pathname notation

Trang 11

The following direct map entry specifies that the client mounts the

/usr/share/mandirectory as read-only from the servers server3,

server4, orserver5, as available

/usr/share/man -ro server3,server4,server5:/usr/share/man

This entry uses a special notation, a comma-separated list of servers, tospecify a powerfulautomountfeature—multiple locations for a fileresource The automountddaemon automatically mounts the

/usr/share/mandirectory as needed, from serversserver3,server4, or

server5, with server proximity and administrator-defined weightsdetermining server selection If the nearest server fails to respond withinthe specified time-out period, the next server with the nearest proximity isselected

Note – Selection criteria for multiple servers, such as server proximity and

administrator-defined weights, is defined in the “Replicated File Systems”section of the automountman page

Adding Indirect Map Entries

The /homeentries define mount points for indirect maps The map

auto_homelists relative path names only Indirect maps obtain the initialpath of the mount point from the master map

#

+auto_master

Trang 12

Creating an Indirect Map

Use theauto_homeindirect map to list the location of home directoriesacross the network For example,

/etc/nsswitch.conffile specifies that NIS, NIS+, or LDAP should besearched

The following describes the syntax for indirect maps:

key [ mount-options ] location

where:

key Specifies the path name of the mount point relative to

the beginning of the path name specified in themaster map

mount-options Specifies the options for a given entry

location Specifies the location of the file resource specified in

server:pathname notation

Trang 13

The example /etc/auto_homefile implies the following mount points:

/home/stevenu,/home/johnnyd,/home/wkd, and /home/mary

Figure 7-4 shows the /home/marymount point

Figure 7-4 The Mount Points

Reducing the auto_home Map to a Single Line

The following entry reduces theauto_homefile to a single line The use ofsubstitution characters specifies that for every login ID, the client

remotely mounts the /export/home/loginIDdirectory from the NFSserver server1onto the local mount point/home/loginID

Figure 7-5 Mounting a Directory on a Local Mount Point

Figure 7-5 shows that this entry uses the wildcard character (*) to matchany key The substitution character (&) at the end of the location is

replaced with the matched key field Using wildcard and substitutioncharacters works only when all home directories are on a single server (in

auto_home autofs

Trang 14

Updating the Automount Maps

When making changes to the master map or creating a direct map, run the

automountcommand to make the changes effective

Running the automount Command

The syntax of the command is:

automount [-t duration] [-v]

where:

You can modify the master map entries or add entries for new maps.However, you must run theautomountcommand to make these changeseffective

You do not have to stop and restart theautomountddaemon after makingchanges to existing entries in a direct map, because the daemon is

stateless You can modify existing entries in the direct map at any time.The new information is used when theautomountddaemon next accessesthe map entry to perform a mount

Any modifications to indirect maps are automatically used by the

automountddaemon

A modification is a change to options or resources A change to the key(the mount point) or a completely new line is an added entry, a deletedentry, or both

-t duration Specifies a time, in seconds, that the file system

remains mounted when not in use The default is

600 seconds (10 minutes)

-v Specifies verbose mode, which displays output as

theautomount command executes

Trang 15

Use Table 7-1 to determine whether you should run (or rerun) the

automountcommand

Note – You can run the automountcommand at any time to rescan themaps, even if running the command is not required

Verifying AutoFS Entries in the /etc/mnttab File

The/etc/mnttabfile is a file system that provides read-only access to thetable of mounted file systems for the current host Mounting a file systemadds an entry to this table Unmounting a file system removes the entryfrom this table Each entry in the table is a line of fields separated byspaces in the form of:

special mount_point fstype options time

where:

Table 7-1 When to Run theautomountCommand

Automount Map Run if the Entry is

Added or Deleted

Run if the Entry

is Modified

special The name of the resource to be mounted

mount_point The path name of the directory on which the file

system is mounted

fstype The type of file system

options The mount options

time The time at which the file system was mounted

Trang 16

You can display the/etc/mnttabfile to obtain a snapshot of the mountedfile systems, including those mounted as an AutoFS file system type.

# cat /etc/mnttab

/dev/dsk/c0t0d0s0 / ufs

rw,intr,largefiles,onerror=panic,suid,dev=2200000 1008255791

/proc /proc proc dev=4080000 1008255790

mnttab /etc/mnttab mntfs dev=4140000 1008255790

Stopping and Starting the Automount System

The /etc/init.d/autofsscript executes automatically as the systemtransitions between run levels, or you can run the script manually fromthe command line

Stopping the Automount System

When theautofsscript runs with the stopargument, it performs aforced unmount of all AutoFS file systems, and it then kills the

automountddaemon

The autofsscript runs with the stopargument automatically whentransitioning to:

● Run level S using the /etc/rcS.d/K41autofsscript

● Run level 1 using the/etc/rc1.d/K41autofsscript

● Run level 0 using/etc/rc0.d/K41autofsscripts script

Trang 17

To run the script on demand, become superuser, and kill theautomountd

daemon by typing the following command:

# /etc/init.d/autofs stop

Starting the Automount System

When the autofsscript is run with thestartargument, the script startsthe automountddaemon, and then it runs the automountutility as abackground task

The script runs with the startargument automatically whentransitioning to run level 2 using the /etc/rc2.d/S74autofsscript

To run the script on demand, become superuser, and start theautomountd

daemon by performing the command:

# /etc/init.d/autofs start

Trang 18

Performing the Exercises

You have the option to complete any one of three versions of a lab Todecide which option to choose, consult the following descriptions of thelevels:

● Level 1 – This version of the lab provides the least amount ofguidance Each bulleted paragraph provides a task description, butyou must determine your own way of accomplishing each task

● Level 2 – This version of the lab provides more guidance Althougheach step describes what you should do, you must determine whichcommands (and options) to input

● Level 3 – This version of the lab is the easiest to accomplish becauseeach step provides exactly what you should input to the system Thislevel also includes the task solutions for all three levels

Trang 19

Exercise: Using the Automount Facility (Level 1)

In this exercise, you use the automount facility to automatically mountman pages and to mount a user’s home directory

Preparation

Choose a partner for this lab, and determine which system will beconfigured as the NFS server and which will serve as the NFS client.Verify that entries for both systems exist in the /etc/hostsfile of eachsystem Refer to the lecture notes as necessary to perform the steps listed.Tasks

Perform the following tasks:

● On the server, perform the steps required to share the

/usr/share/mandirectory

● On the client, rename the /usr/share/mandirectory to

/usr/share/man.origdirectory, and create a new mount point forthe/usr/share/mandirectory Edit the master map so that it calls adirect map Create the direct map to mount the/usr/share/man

directory from the server Use theautomountcommand to updatetheautomountddaemon Test that the man pages work, and verifythe mount that occurs

● Create a new, identical user on both the server and client that uses

/export/home/usernamefor the user’s home directory On bothsystems, make the changes required in the/etc/passwdfile to setthe home directory for this new user to the/home/username

directory

● On the server, perform the steps required to share the/export/home

directory

● On both systems, make the changes required in the/etc/auto_home

file to allow both systems to automatically mount the

/export/home/usernamedirectory when the new user calls for the

/home/usernamedirectory Test the new user login on both systems,and verify that the mounts take place Log in asrootwhen finished

● On the server, unshare the /export/homeand/usr/share/man

directories, and remove entries for these directories from the

/etc/dfs/dfstabfile Stop the NFS server daemons

● On the client, remove the direct map entry from the

file, and update the daemon with

Trang 20

Preparation

Choose a partner for this lab, and determine which system will beconfigured as the NFS server and which will serve as the NFS client.Verify that entries for both systems exist in the/etc/hostsfile of eachsystem Refer to the lecture notes as necessary to perform the steps listed.Task Summary

● On the client, rename the/usr/share/mandirectory to

/usr/share/man.origdirectory, and create a new mount point forthe/usr/share/mandirectory Edit the master map so that it calls adirect map Create the direct map to mount/usr/share/man

directory from the server Use theautomountcommand to updatetheautomountddaemon Test that the man pages work, and verifythe mount that occurs

directory

/home/usernamedirectory Test the new user login on both systems,and verify that the mounts take place Log in as rootwhen finished

● On the server, unshare the /export/homeand /usr/share/man

directories and remove entries for these directories from the

Trang 21

Complete the following tasks

Task 1– On the Server Host

Complete the following steps:

1 Edit the/etc/dfs/dfstabfile, and add a line to share the manpages

2 Use the pgrepcommand to check if themountddaemon is running

● If the mountddaemon is not running, start it

● If the mountddaemon is running, share the new directory

Task 2 – On the Client Host

1 Rename the/usr/share/mandirectory so that you cannot view theman pages installed on the client system

4 Run the automountcommand to update the list of directories

managed by theautomountddaemon

_

5 Test the configuration, and verify that a mount for the

/usr/share/mandirectory exists after accessing the man pages. _What did you observe to indicate that theautomountoperation wassuccessful?

_

Trang 22

Task 3 – On the Server Host

1 Verify that the/export/homedirectory exists If it does not exist,create it

2 Add a user account with the following characteristics:

● User ID: 3001

● Primary group: 10

● Home directory: /export/home/usera

● Login shell:/bin/ksh

● User name: usera

3 Remove the lock string from the new user’s/etc/shadowfile entry

● User ID: 3001

● Login shell: /bin/ksh

Trang 23

Task 5 – On Both Systems

1 Edit the/etc/passwdfile, and change the home directory for thenew user from the/export/home/usernamedirectory to

/home/username, where usernameis the name of your new user

2 Edit the /etc/auto_homefile Add the following line, and replace

usernamewith the name of your new user:

username server:/export/home/username

1 Edit the/etc/dfs/dfstabfile, and add a line to share the

/export/homedirectory

2 Use the pgrepcommand to check if themountddaemon is running

● If the mountddaemon is not running, start it

● If the mountddaemon is running, share the new directory

Complete the following step:

Log in as the new user

Do both systems automatically mount the new user’s homedirectory?

_Which directory is mounted, and what is the mount point:

● On the server?

● On the client?

Trang 24

1 On the client, log off asusera. _

2 Remove the entry for userafrom the/etc/auto_homemap

1 Log off asusera. _

2 After the client reboots as described in Step 4 of ‘‘Task 8 – On theClient Host’’ section on page 7-24, remove the entry for userafromthe/etc/auto_homemap

Trang 25

Preparation

Choose a partner for this lab, and determine which system will beconfigured as the NFS server and which will serve as the NFS client.Verify that entries for both systems exist in the /etc/hostsfile of eachsystem Refer to the lecture notes as necessary to perform the steps listed

Task Summary

● On the client, rename the /usr/share/mandirectory to

/usr/share/man.origdirectory, and create a new mount point forthe/usr/share/mandirectory Edit the master map so that it calls adirect map Create the direct map to mount/usr/share/man

directory from the server Use theautomountcommand to updatetheautomountd daemon Test that the man pages work, and verifythe mount that occurs

directory

/home/usernamedirectory Test the new user log in on both systems,and verify that the mounts that happen Log in asrootwhen

finished

Trang 26

● On the server, unshare the /export/homeand /usr/share/man

directories and remove entries for these directories from the

/etc/dfs/dfstabfile Stop the NFS server daemons

● On the client, remove the direct map entry from the

/etc/auto_masterfile, and update theautomountddaemon withthe change Return the /usr/share/mandirectory to its originalconfiguration

Tasks and Solutions

The following section provides the tasks with their solutions

1 Edit the /etc/dfs/dfstabfile, and add a line to share the manpages

1 Rename the/usr/share/mandirectory so that you cannot view theman pages installed on the client system

# cd /usr/share/

# mv man man.orig

Trang 27

2 Edit the /etc/auto_masterfile to add an entry for a direct map.

5 Test the configuration, and verify that a mount for the

/usr/share/mandirectory exists after accessing the man pages

# man ls

< output from man command >

# mount | grep man

/usr/share/man on sys44:/usr/share/man

remote/read/write/setuid/dev=42c0003 on Thu Dec 13 08:07:26 2001

What did you observe to indicate that theautomountoperation wassuccessful?

This operation should automatically mount the directory in which the manuals are stored In other words, themancommand should work.

# ls /export/home

Note – Perform the next command if the /export/homedirectory doesnot exist

# mkdir /export/home

Trang 28

● User ID: 3001

● Login shell:/bin/ksh

# useradd -u 3001 -g 10 -m -d /export/home/usera -s /bin/ksh usera

1 Verify that the/export/homedirectory exists If it does not, create it

● Login shell: /bin/ksh

# useradd -u 3001 -g 10 -m -d /home/usera -s /bin/ksh usera

1 Edit the /etc/passwdfile, and change the home directory for thenew user from the/export/home/usernamedirectory to

/home/username, whereusernameis the name of your new user

# vi /etc/passwd

Trang 29

1 Edit the/etc/dfs/dfstabfile, and add a line to share the

Complete the following step:

Log in as the new user

# su - usera

Do both systems automatically mount the new user’s homedirectory?

Yes, this should work.

Which directory is mounted, and what is the mount point:

Trang 30

1 On the client, log off asusera

2 Remove the entry for userafrom the/etc/auto_homemap

3 Remove the entry for theauto_homemap from the

1 Log off asusera

2 After the client reboots as described in Step 4 of ‘‘Task 8 – On theClient Host’’ section on page 7-30, remove the entry for userafromthe/etc/auto_homemap

3 Remove the entries from /etc/dfs/dfstabfile

4 Unshare mounted directories

# unshareall

Trang 31

Exercise Summary

?

!

Discussion – Take a few minutes to discuss what experiences, issues, or

discoveries you had during the lab exercise

● Experiences

● Interpretations

● Conclusions

● Applications

Trang 33

Describing RAID and the Solaris™ Volume

Manager Software

Objectives

A redundant array of independent disks (RAID) configuration enablesyou to expand the characteristics of a storage volume beyond the physicallimitations of a single disk You can use a RAID configuration to increasedisk capacity as well as to improve disk performance and fault tolerance.The Solaris™ Volume Manager software provides a graphical user

interface (GUI) tool to simplify system administration tasks on storagedevices Upon completion of this module, you should be able to:

● Describe RAID

● Describe Solaris Volume Manager software concepts

The following course map shows how this module fits into the currentinstructional goal

Figure 8-1 Course Map

Describing RAID and Solaris

Volume Manager Software

Managing Storage Volumes

Configuring Solaris Volume Manager Software

Trang 34

Introducing RAID

RAID is a classification of methods to back up and to store data onmultiple disk drives There are six levels of RAID as well as anon-redundant array of independent disks (RAID 0) The Solaris VolumeManager software uses metadevices, which are product-specific

definitions of logical storage volumes, to implement RAID 0, RAID 1,and RAID 5:

● RAID 0: Non-redundant disk array (concatenation and striping)

● RAID 1: Mirrored disk array

● RAID 5: Block-interleaved distributed-parity

RAID 0

RAID-0 volumes, including both stripes and concatenations, arecomposed of slices and let you expand disk storage capacity You caneither use RAID-0 volumes directly or use the volumes as the buildingblocks for RAID-1 volumes (mirrors) There are two types of RAID-0volumes:

● Concatenated volumes (or concatenations)

A concatenated volume writes data to the first available slice Whenthe first slice is full, the volume writes data to the next available slice

● Striped volumes (or stripes)

A stripe distributes data equally across all slices in the stripe

RAID-0 volumes allow you to expand disk storage capacity efficiently.These volumes do not provide data redundancy If a single slice fails on aRAID-0 volume, there is a loss of data

Trang 35

Concatenated Volumes

Figure 8-2 shows that in a concatenated RAID 0 volume, data is organizedserially and adjacently across disk slices, forming one logical storage unit

Figure 8-2 RAID-0 Concatenation

A concatenation combines the capacities of several slices to get a largerstorage capacity You can add more slices to the concatenation as thedemand for storage increases You can add slices at anytime, even if otherslices are currently active

Note – An interlace is a grouped segment of blocks on a particular slice.

Interlace 2 Interlace 3 Interlace 4

Physical

Slice C

Interlace 5

Solaris Volume Manager

Interlace 12

Interlace 2 Interlace 1

Interlace 10 Interlace 11 Interlace 12 Interlace 9

Trang 36

The default behavior of concatenated RAID-0 volumes is to fill a physicalcomponent within the volume before beginning to store data on

subsequent components within the concatenated volume However, thedefault behavior of UFS file systems within the Solaris OE is to distributethe load across devices assigned to the volume containing a file system.This anomaly can make it seem that concatenated RAID-0 volumesdistribute data across the components of the volume in a round-robinmethod The data distribution is a function of the UFS file system that ismounted in the concatenated volume and is not a function of the

concatenated volume itself

You can also use a concatenation to expand any active and mounted UFSfile system without having to bring down the system Usually, the

capacity of a concatenation is the total size of all the slices in theconcatenation

Trang 37

Striped Volumes

Figure 8-3 shows the arrangement of a RAID-0 volume A RAID 0 volumeconfigured as a stripe arranges data across two or more slices Stripingalternates equally-sized segments of data across two or more slices,

forming one logical storage unit These segments are interleaved

round-robin, so that the combined space is created alternately from eachslice

Figure 8-3 RAID-0 Stripe

Striping enables parallel data access because multiple controllers canaccess the data at the same time Parallel access increases Input/Output(I/O) throughput because multiple disks in the volume are busy servicingI/O requests simultaneously

Physical

Slice A PhysicalSlice B PhysicalSlice C

Solaris Volume Manager

RAID 0 (Stripe) Logical Volume

Interlace 4 Interlace 5

Interlace 2

Trang 38

You cannot convert an existing file system directly to a stripe You mustfirst back up the file system, create the stripe, and then restore the filesystem to the stripe.

For sequential I/O operations on a stripe, the Solaris Volume Manager

software reads all the blocks in an interlace An interlace is a grouped

segment of blocks on a particular slice The Solaris Volume Managersoftware then reads all the blocks in the interlace on the second slice, and

so on

An interlace is the size, in Kbytes, Mbytes, or blocks, of the logical datachunks on a stripe Depending on the application, different interlacevalues can increase performance for your configuration The performanceincrease comes from several disk head-arm assemblies (HDAs)

concurrently executing I/O operations When the I/O request is largerthan the interlace size, you might get better performance

When you create a stripe, you can set the interlace value or use the SolarisVolume Manager software’s default interlace value of 16 Kbytes Afteryou create the stripe, you cannot change the interlace value (although youcould back up the data on it, delete the stripe, create a new stripe with anew interlace value, and then restore the data)

Trang 39

RAID 1

RAID-1 volumes, also known as mirror volumes, are typically composed

of RAID-0 volumes and provide the advantage of data redundancy Thedisadvantage is the higher cost incurred by requiring two RAID-1 deviceswherever a single RAID-0 device is mirrored Typical topics to be

considered when configuring mirrors are:

● Trade-offs when using mirrors

● Uses of multiple submirrors

● RAID 0+1

● RAID 1+0

● Mirror read, write, and synchronization options

● Mirror configuration guidelines

Trade-Offs When Using Mirrors

A RAID-1 (mirror) volume maintains identical copies of the data in

RAID-0 volumes Mirroring requires more disks You need at least twice

as much disk space as the amount of data to be mirrored

After configuring a mirror, you can use it as if it were a physical slice.With multiple copies of data available, data access time is reduced if themirror read and write policies are properly configured You then use readand write policies to distribute the access to the submirrors evenly acrossthe mirror The mirror read and write policies are described in detail later

in this module

You can mirror any file system, including existing file systems You canalso use a mirror for any application, such as a database

Using Multiple Submirrors

A mirror is made of two or more RAID-0 volumes configured as eitherstripes or concatenations The mirrored RAID-0 volumes are called

submirrors A mirror consisting of two submirrors is known as a two-waymirror, while a mirror consisting of three submirrors is known as a

three-way mirror

Creating a two-way mirror is usually sufficient for data redundancy A

Trang 40

When a submirror is offline, it is in a read-only mode The Solaris VolumeManager software tracks all the changes written to the online submirror.When the submirror is brought back online, only the newly writtenportions are resynchronized Other reasons for taking the submirroroffline include troubleshooting and repair.

You can attach or detach a submirror from a mirror at any time, though atleast one submirror must remain attached to the mirror at all times.Usually, you begin the creation of a mirror with only a single submirror,after which you can attach additional submirrors

Figure 8-4 RAID-1 Mirror

The Solaris Volume Manager software makes duplicate copies of the datalocated on multiple physical disks The Solaris Volume Manager softwarepresents one virtual disk to the application All disk writes are duplicated,and disk reads come from one of the underlying submirrors If the

submirrors are not of equal size, the total capacity of the mirror is limited

by the size of the smallest submirror

Interlace 2 Interlace 3 Interlace 4 Interlace 1

Submirror 1

RAID 1 (Mirror) Logical Volume

Submirror 2 Submirror 1

Submirror 2 Solaris VolumeManager

Int 1 Int 2 Int 3 Int 4

Định dạng
Số trang	86
Dung lượng	842,83 KB