[Previous Section] [Back to Table of Contents] [Next Section]

IRIX Advanced Site and Server Administration Guide

Chapter 2
Operating the System

This chapter deals with the day-to-day operations of an IRIS workstation or server running the IRIX Operating System. It covers:

Starting the System

To start up a system, follow these steps:

    Make sure all cables (such as power, display monitor cable, and keyboard) are properly connected. See your owner's guide and hardware guide for complete information about cabling your particular workstation or server.

    Turn on the power switches on the display monitor (or console terminal) and the computer.

    The computer runs power-on diagnostics and displays some copyright messages and some system startup information. These messages appear on the console screen or on the screen of a diagnostics terminal (an ASCII terminal connected to the first serial port) of a server. A copy of these messages is also written to the /var/adm/SYSLOG file in case you miss them.

    If the operating system determines that the file systems need checking, it checks them with the fsck program. fsck fixes any problems it finds before the operating system mounts the file systems. fsck will run if the system is not shut down properly, such as in the event of a power failure. For information about using fsck, see "Maintaining File Systems" in this guide, and the fsck(1M) reference page. Note that it is not necessarily a problem if fsck runs, it is merely a precaution.

    The system now comes up in multiuser mode and you can log in. You should leave your system running at all times. The IRIX operating system works best when it is allowed to run continuously, so that scheduled operations and ``housekeeping'' processes can be performed on schedule.

Shutting Down the System

This section describes how to turn off a workstation or server from multiuser or single-user mode.

Shutting Down from Multiuser Mode

To shut down the system from multiuser mode:

    Use the who(1) command to determine which users are logged in to the operating system, if any: 

    who

    Notify any users that the system is shutting down. Issue the /etc/wall(1M) command:

    wall

    Enter your message. For example, you might enter:

    There is a problem with the building's power system. 
I will be shutting down the system in 10 minutes. 
Please clean up and log off. 
Sorry for the inconvenience, 
norton

    When you finish entering your message, type <Ctrl-D>. The message is sent to all users on the system. They see something like this:

    Broadcast Message from root Tue Oct 17 17:02:27...
There is a problem with the building's power system. 
I will be shutting down the system in 10 minutes.
Please clean up and log off.
Sorry for the inconvenience,
norton

    Issue the /etc/shutdown command:

    /etc/shutdown -y -i0 -g600

    The above command specifies a 10 minute (600 second) grace period to allow users to clean up and log off. The other flags indicate that the system is to be completely shut down ( -i0) and that the system can assume that all answers to any prompts regarding the shutdown are ``yes.'' You see the following message:

    Shutdown started. Fri Aug 28 17:10:57...
Broadcast Message from root (console) Fri Aug 28 17:10:59
The system will be shut down in 600 seconds. 
Please log off now.

    After ten minutes, you see this message:

    INIT: New run level: 0
The system is coming down. Please wait.

    The Command Monitor prompt or System Maintenance menu appears. Wait for a Command Monitor prompt or System Maintenance menu to appear before turning off power to the workstation or you may damage your hard disk.

    You can now turn off the power.

For more information on shutting down the system, see the halt(1M) and shutdown(1M) reference pages. Remember that you should shut down the system only when something is wrong or if modifications to the software or hardware are necessary. IRIX is designed to run continuously, even when no users are logged in and the system is not in use.

Turning Off from Single-user Mode

If the system is in single-user mode, follow these steps:

    Use the shutdown command to turn off the system and guarantee file system integrity. As root, enter the command:

    shutdown -y -i0 -g0

    where:

    -y assumes yes answers to all questions, -i0 goes to state 0 (System Maintenance Menu), and -g0 allows a grace period of 0 seconds.

    You see a display similar to this one:

    Shutdown started. Fri Aug 28 17:11:50 EDT 1987
INIT: New run level: 0
The system is coming down. Please wait.

    The system stops all services and the Command Monitor prompt or System Maintenance Menu appears.

    Wait for the Command Monitor prompt or System Maintenance menu to appear or for a message that the system can be powered off before turning off power to the computer. Doing so prematurely may damage your hard disk.

    Turn off power to the computer.

Using Regular Expressions and Metacharacters

There are shortcuts available to you when you wish to define large numbers of files or directories in your IRIX shell commands. These shortcuts are known as "regular expressions." Regular expressions are made up of a combination of alpha-numeric characters and a series of punctuation characters that have special meaning to the IRIX shells. These punctuation characters are called metacharacters when they are used for their special meanings with shell commands. The following is a list of the IRIX metacharacters:

Table 2-1 : IRIX Metacharacters
Metacharacter Meaning
* wildcard
? single character wildcard
[] set definition marks



The asterisk (*) metacharacter is a universal wildcard. This means that the shell interprets the character to mean any and all files. For example, the command:

cat *

tells the shell to concatenate all the files in a directory, in alphabetical order by filename. The command:

rm *

tells the shell to remove everything in the directory. Only files will be removed, though, since a different command, rmdir(1) is used to remove directories. However, the asterisk character does not always have to refer to whole files. It can be used to denote parts of files as well. For example, the command:

rm *.old

will remove all files with the suffix.old on their names.

The single character wildcard is a question mark (?). This metacharacter is used to denote a wildcard character in one position. For example, if your directory contains the following files:

file1 
file2 
file3 
file.different

and you wish to remove file1, file2, and file3, but not file.different, you would use the command:

rm file?

If you used an asterisk in place of the question mark, all your files would be removed, but since the question mark is a wildcard for a single space, your desired file is not chosen.

Square brackets denote members of a set. For example, consider the list of files used in the example of the single character wildcard. If you wanted to remove file1 and file2, but not file3 or file.different, you would use the following command:

rm file[12]

This command tells the shell to remove any files with names starting with file and with the character 1 or 2 following, and no other characters in the name. Each character in the brackets is taken separately. Thus, if our example directory had included a file named file12, it would not have been removed by the above command. You can also use a dash (-) to indicate a span of characters. For example, to remove file1, file2, and file3, use the following command:

rm file[1-3]

Alphabet characters can be spanned as well, in alphabetical order. The shell does pay attention to upper case and lower case letter, though, so to select all alphabet characters within square brackets, use the following syntax:

[a-z,A-Z]

You can use the square brackets in combination with other metacharacters as well. For example, the command:

rm *[2,3]

removes any files with names ending with a 2 or 3, but not file1 or file.different.

Using Shell Shortcuts in the IRIX System

IRIX provides a number of facilities that can be used for your convenience as you administer and use the system. Some of these shortcuts are available through the command shells (/usr/bin/csh, /usr/bin/tcsh, /usr/bin/sh and /usr/bin/ksh) and others are made available through the computer hardware itself. These shortcuts are useful because they minimize keystrokes. While minimizing keystrokes may seem to be a minor concern at first glance, an administrator who issues lengthy and complex command lines repeatedly may find these shortcuts a handy and necessary time-saving feature.

C Shell

The IRIX C Shell (/bin/csh) provides several features that can be used to minimize keystrokes for routine tasks. Complete information about these and many other features of the C Shell is available in the csh(1) reference page. Among the features provided are:

filename completion


This feature is activated with the command:

set filec

Filename completion allows you to enter the first character or two of a command or file name and then press the Escape key to have the shell complete the name. This is useful when you have long filenames with many suffixes. If more than one file or directory or command matches the characters you have given, the shell completes as much as possible of the name, and then prompts you with a beep for more information. You can also use the <Ctrl-D> character to select all files or directories that match your given characters.

shell scripts

This feature allows you to create a program that will be executed by the shell. This feature is similar to a programming language in that it has a set syntax and set of instructions, yet it requires no compiler and produces no object file; it is directly executed by the shell. Many administrators use this feature for frequently performed procedures that require some planning and complex execution, such as finding large files and notifying the owners that such files cannot be kept on the system for long periods of time. The shell script programming rules are clearly presented on the csh(1) reference page.

input/output redirection


This feature allows you to direct the output of a command into a file or into another command as input. You can also direct a command to take its input from a file. It is often used as part of a shell script, but is generally used on the command line to string together a series of commands. For example, consider the command line:

ps -ef | grepcommandname

The pipe character directs the shell to use the output of the ps command as the input to the grep command. The result is that all instances of the command commandname in the process list are printed on the screen, saving the administrator the effort of searching through the process listing.

job control

This feature allows you to use a single screen (or shell window) to manage several programs running simultaneously. It is most useful for the server administrator who manages the system from a single character-based terminal.

command aliasing


This feature allows you to create aliases for commonly used command strings, saving keystrokes. For example, suppose you frequently give the command:

ls -CF | more

This command line executes the ls command with certain options and ensures that if the output is greater than a screenful it will be stopped until you have read it. However, it would be tedious to type the whole command each time you wanted to see a directory listing in your preferred format. Therefore, you should create an alias. You can alias the above command line to any series of keystrokes you like. You can even alias it to "ls," thus bypassing the standard meaning of the ls command.

When you create the alias, however, be aware that any command that requires one or more arguments, or one such as ls that may or may not receive arguments, must have a provision made in the alias for those arguments. The standard provision made in aliases for possible arguments is the following regular expression:

\!*

The leading backslash escapes the initial meaning of the exclamation point to the shell and passes the exclamation point through to the command line, where it is interpreted by the shell to refer to arguments given on the aliased command line. The asterisk in the expression means that all characters typed in as arguments are to be passed through to the shell. As an example, the line you place in your .cshrc file to create the example alias is:

alias ls "ls -CF \!* | more"

Then, when you type the command:

ls filename

at your shell prompt, the command is executed as:

ls -CF filename | more

Aliases can be used freely within shell scripts, with filename completion and full use of regular expressions and output redirection.

command history


The shell maintains a log of your past commands given during this login session. You can repeat or edit a previously given command to save keystrokes. The history command shows the numbered log of commands in order. The first command given in your login session is number 1, the second is number 2, and so on. You can set the number of commands the shell remembers in your .cshrc file. To execute the most recent command again, type:

!!

To execute the most recent command beginning with the letter "q," use the command line:

!q

And to execute a command by its number in the history, give the command line:

!n

where n is the number of the previous command you wish to re-execute.

Tcsh Shell

The /usr/bin/tcsh program is an improved version of the C shell. In addition to the C shell features listed above, this shell offers many new features. A few of the most useful to system administrators are:

There are many more features implemented in Tcsh, and all of them are covered in the tcsh(1) reference page.

Bourne Shell

The Bourne shell (/bin/sh) provides fewer features than the C shell, but in its place offers a level of access to the shell that contains far fewer restrictions and intervening layers of interface. For example, you can write shell script programs directly from the shell prompt with Bourne shell. Input and output redirection and command aliasing are supported with the Bourne shell, but no command history, job control, or filename completions are available. For a complete discussion of the Bourne shell and its features, see the sh(1) reference page.

Korn Shell

The Korn shell was developed to provide the best features of both the C shell and the Bourne shell. The /bin/ksh program provides the ease of shell programming found in the Bourne shell, along with the job control, history mechanism, filename completion, and other features found in the C shell. This shell has changed many of the ways these features are implemented, and also provides improved command line editing facilities. See the ksh(1) reference page for complete information on this shell. Useful features include:

Emacs Editing

This mode is entered by enabling either the emacs or gmacs option. To edit, the user moves the cursor to the point needing correction and then inserts or deletes characters or words as needed as if the command line were a text file being edited using Emacs. All edit commands operate from any place on the line (not just at the beginning).

vi Editing

To enter this mode, enable the vi option. There are two typing modes in this option. Initially, when you enter a command you are in the input mode. To edit, the user enters control mode by typing ESC, moves the cursor to the point needing correction, then inserts or deletes characters or words as needed as if the command line were a text file being edited using vi.

job control

Lists information about each given process ( job) or all active processes if the job argument is omitted. The -l flag lists process ID numbers in addition to the normal information. The -n flag only displays jobs that have stopped or exited since last notified. The -p flag causes only the process group to be listed. See the ksh(1) reference page for a description of the format of the job argument.

The bg command puts each specified process into the background. The current process is put in the background if job is not specified.

The fg command brings each process specified to the foreground. Otherwise, the current process is brought into the foreground.

Mouse Shortcuts

The system hardware for graphical workstations (and some X-terminals) can provide you with shortcuts. These may not be available to server administrators who rely solely on character-based terminals for their administration. Using the graphics console of your system, you can cut and paste between windows without using pull-down or pop-up menus of any sort. Using the pop-up menu, you can manipulate your windows completely.

Note that you can customize the action of your mouse buttons. All examples in this section assume the default mouse button meanings are being used. If you have modified your mouse action, you must allow for that modification before you use these techniques.

For complete information on using the pop-up windows, see your IRIS Essentials book, either in hard copy or on screen through the IRIS InSight software package.

Using the Mouse to Copy and Paste Text

The most common mouse shortcut is to cut, copy, and paste between windows on your screen. Here is how you do it:

    Find the cursor controlled by your mouse on your screen. It should appear as a small arrow when it is positioned in the working area of one of your windows, or as an "X" when it is positioned on your background screen, or as some other figure when it is positioned on the frame of a window or in the working area of an application's window. If you can't locate the cursor immediately, move the mouse around a bit and look for motion on your screen. You should find the cursor easily.

    Place the cursor at the beginning of the text you wish to paste between windows and press the leftmost key on the top of the mouse. Now, keeping the mouse button depressed, move the cursor to the end of the text you wish to paste. The intervening area of the window changes color to show the selected text. If you are selecting a large section of text, it is not necessary to move the cursor over every space. You may move the cursor directly to the end point and all intervening text will be selected. It is not possible to select "columns" of text or several disconnected pieces of text at once. When you have moved the cursor to the desired end point, release the mouse button. The text remains highlighted.

    Now move the cursor to the window you want to paste the text into and make certain the window is ready to receive the pasted text. For example, if you are pasting a long command line, make certain that there is a shell prompt waiting with no other command already typed in. If the pasted matter is text into a file, make certain that the receiving file has been opened with an editor and that the editor is in a mode where text can be inserted.

    To paste the text, place the cursor in the receiving window and press the middle mouse button once quickly. Each time you press the middle button, the selected text will be pasted into the receiving window. Sometimes it takes a moment for the text to appear, so be patient. If you press the button several times before the text appears, you will paste several copies of your text.

    You can also paste your selected text to the bottom of a window (including the window from which you selected the text). Press the rightmost mouse button while the cursor is in that window and select the send option from the pop-up menu that appears.

The text you originally selected remains selected until you select new text somewhere else or until you place the cursor back in the original window and click once on the leftmost mouse button.

Using the Mouse to Create a New Shell Window

If you need a new shell window, you can use the mouse to create one. Follow these steps:

    With the cursor in a shell window, press the rightmost button on your mouse. A pop-up menu appears:

    ch02.popup.gif

    Figure 2-1 : Shell Pop-Up Menu


    The last item on the pop-up menu is the clone option. There is a small triangle to the right of this option. This triangle indicates that there are more sub-choices available in another pop-up menu. While keeping the button on the mouse depressed, move the mouse down until the clone option is highlighted and the sub-menu pops up, showing various shell window cloning options. These options create another shell window functionally identical to the one in use. This is why the option is called cloning. The text and background colors of the current window are carried forward to the cloned window, and the selections in the sub-menu specify the number of lines in the new window. You can choose to have the same number of lines in the cloned window as in the current window, or to have 24, 40, or 60 lines.

    ch02.custom.gif

    Figure 2-2 : Shell Window Cloning Submenu


    Select the size you want by moving the mouse down to highlight each option and releasing the mouse button when the option you desire is highlighted. The new window will appear on your screen presently. You may repeat this process as often as you like on any shell window.

Creating a Custom Shell Window

IRIX allows you to create a shell window using any colors you like from the palette on your graphics workstation. You may also select any font you prefer from the font set on your system. The xwsh(1) command creates the shell window, and the options to this command control the various fonts, colors, and other features available to you.

For a complete list of the features available with xwsh(1), see the xwsh reference page. The most commonly used features are described here in the following examples.

To create a simple shell window with a dark gray background and yellow text, issue the following command:

xwsh -fg yellow -bg gray40 &

The above command generates a new window and a new shell using the colors specified. The window will use the default font selection and window size, since these attributes were not specified. The command that created the shell was placed in the background, so the shell does not tie up the window where you gave the command. You can always place a command in the background by adding the ampersand character (&) to the end of the command line. For more information on placing processes in the background, see the csh(1) reference page.

There are 100 shades of gray available. Gray0 is the darkest, and is virtually black. Gray100 is the lightest and is virtually white. The effect of selecting foreground (text) in yellow and background in gray40 is similar to yellow chalk on a gray chalkboard. For a complete list of the available colors in your palette, use the colorview(1) command. This brings up a window with the list of colors in a scrollable list, and a display window to show a patch of the currently selected color.

In the next example, we change the colors to black on a sky blue background (high contrast between the foreground and background makes reading the screen easier), and we add a specification for the size of the window. 

xwsh -fg black -bg skyblue -geometry 80x40 &

The first number in the geometry option is 80, indicating that the new shell window should be 80 characters wide (this is the default). The second number indicates the desired number of lines on the screen, in this case 40. Once again, the xwsh command has been placed in the background by adding the ampersand character to the end of the command line.

You can make your new shell come up on your desktop as an icon by adding the -iconic flag to any xwsh command.

To select a font other than the default, you can use the on-screen font selection utility, or you can specify the font on the command line. It is a great deal easier to use the on-screen font selection utility, as you must specify a great number of attributes for the font on the command line. Also, it frequently takes a great number of selections before you settle on a font, a weight (regular or bold, condensed or normal) and a font size that appeals to you. Using the on-screen font utility, you can preview what each selection will look like on your windows.

Once you have made your selections, you can copy and paste the font selection information and the rest of your xwsh command into a shell script file for convenient future use. For example, here is an xwsh command line that specifies the IRIS-specific font haebfix in a medium weight with normal spacing, 15 pixels tall. The remaining information is generated by the font selection utility for the shell. 

xwsh -iconic -fg yellow -bg grey40 -geometry 80x40 -fn \ -sgi-haebfix-medium-r-normal--15-150-72-72-m-90-iso8859-1 &

Note that in your shell script, the above command appears all on one line. Due to formatting constraints, the command is broken across two lines in this example.

For complete information on using the font selection utility in xwsh and the xfontsel(1) command, see Chapter 2 of the IRIS Utilities Guide.

Finding and Manipulating Files Automatically

The IRIX system provides several tools for manipulating large numbers of files quickly. Some of the most common are described below:

Using find to Locate Files

The find(1) command is used to find files and possibly execute commands on the files found. It starts at a given directory and searches all directories below the starting directory for the specified files. A basic find command line looks like this:

find . -name 


<filename> -print

This command searches from the current directory until it finds the exact <filename> and then displays its location on your screen. You can also use regular expressions (see "Using Regular Expressions and Metacharacters") in your searches.

The following command line searches for files that have been changed after the time /tmp/file was last modified. If you use touch(1) to create /tmp/file with an old date, this command can help you find all files changed after that date.

find / -local -newer /tmp/file -print

You can use find to locate files and then to run another command on the found files. This example shows how to locate a file in a user's directory:

cd /usr/people/trixie 
find . -name 'missingfile' -print

In this example, the period (.) indicates the current directory, the -name option indicates that the next argument in quotes is the name of the file you are looking for, and the -print option tells find to display the pathname of the file when the file is located.

The next example shows how to change the permissions on all the files in the current directory and in all subdirectories:

find . -name '*' -local -exec chmod 644 {} \;

The option immediately following the find command is a period (.). This indicates to find that the search is to begin in the current directory and include all directories below the current one. The next flag, -name, indicates the name of the files that are being found. In this case, all files in the directory structure are selected through the use of the asterisk metacharacter (*). See "Using Regular Expressions and Metacharacters" for more information on metacharacters and regular expressions.

The -local option indicates to find that the search is limited to files that physically reside in the directory structure. This eliminates files and directories that are mounted via the Network File System (NFS). The -exec option causes find to execute the next argument as a command, in this case chmod 644. The braces, { }, refer to the current file that find is examining.

The last two characters in the command line are part of the chmod command that will be executed (with the -exec option) on all files that match the search parameters. The backslash (\) is necessary to keep the shell from interpreting the semicolon (;). The semicolon must be passed along to the chmod process. The semicolon indicates a carriage return in the chmod command.

find has several other useful options:

-inum n

Locate files by their inode number (n) instead of their name.

-mtime n

Identify files that haven't been modified within a certain amount of time (n).

-perm [ -] ||onum


Identify files with permissions matching onum, an octal number that specifies file permissions. See the chmod(1) reference page. Without the minus sign (-), only file permissions that match exactly are identified.

If you place a minus sign in front of onum, only the bits that are actually set in onum are compared with the file permission flags.

-type x

Identifies files by type, where x specifies the type. Types can be b, c, d, l, p, f, or s for block special file, character special file, directory, symbolic link, FIFO (a named pipe), plain file, or socket respectively.

-links n

Matches files that have n number of links.

- user uname

Identifies files that belong to the user uname. If uname is a number and does not appear as a login name in the file /etc/passwd, it is interpreted as a user ID.

- group gname

Identifies files that belong to the group gname. If gname is a number and does not appear in the file /etc/group, it is interpreted as a group ID.

- size n [c]

Identifies files that are n blocks long (512 bytes per block). If you place a c after the n, the size is in characters.

- ok cmd

Works like -exec, except a question mark (?) prompts you to indicate whether you want the command (cmd) to operate on the file that is found. This is useful for such operations as selectively removing files.

Copying Directories or Directory Hierarchies

The find and cpio commands can be used to easily and safely copy a directory or a directory hierarchy as long as the user has permissions to access the directory. To copy a directory with all its files, or an entire hierarchy of directories and their files, use commands like the following:

mkdir new_directory_name 
cd the_directory_you_want_to_copy 
find . -print | cpio -pdlmv new_directory_name

This command sequence preserves the symbolic links in the new directory as well as transparently cross file system boundaries.

Automated Editing with sed

You can use sed(1), the Stream Editor, to automate file editing. sed follows an editing script that defines changes to be made to text in a file. sed takes a file (or files), performs the changes as defined in the editing script, and sends the modified file to the standard output. sed is fully described in the IRIX Development Option documentation and in the sed(1) reference page, which is included in your IRIX distribution. The IRIX Development Option is available for separate purchase from Silicon Graphics.

Some Recursive Commands

Recursive commands can save you a lot of time. For example, to change the ownership of all the files and directories in a directory recursively, and all of the files and directories in all of the subdirectories below that, you can use the recursive option with chown(1):

chown -R username directory

Some of the other commands in the IRIX system that have recursive options are:

ls -R 
rm -r 
chgrp -R

If you want to use a particular command recursively, but it does not have a recursive option, you can run the command using find. See "Using find to Locate Files".

Note that using recursive options to commands can be very dangerous in that the command automatically makes changes to your files and file system without prompting you in each case. The chgrp command can also recursively operate up the file system tree as well as down. Unless you are sure that each and every case where the recursive command will perform an action is desired, it is better to perform the actions individually. Similarly, it is good practice to avoid the use of metacharacters (described in "Using Regular Expressions and Metacharacters") in combination with recursive commands.

Automating Tasks with at(1), batch(1), and cron(1M)

You can use the at(1), batch(1), and cron(1M) utilities to automate many of your usual tasks, both as an administrator and as a user. These utilities perform similar functions. All execute commands at a later point in time. The difference between the commands is that at executes the given command at one specific time; cron sets up a schedule and executes the command or commands as often as directed, according to the schedule; and batch executes the commands when system load levels permit the execution.

at(1) Command

If you have a task to process once at a later point in time, use at. For example, if you wish to close down permissions on a public directory at midnight of the current day, but you do not need to be present when this occurs, you could use the command string:

at 2400 July 14 
chmod 000 /usr/public 
<Ctrl-D>

It is required that the at command itself and the date and time of the command be placed alone on a line. When you press <Return>, you do not see a prompt; at is waiting for input. Enter the command to be executed just as you would type it at a shell prompt. After entering the command, press <Return> again and enter <Ctrl-D> to tell at that no more commands are forthcoming. You can use a single at command to execute several commands at the appointed time. For example, if you want to close the public directory and change the message of the day to reflect this closure, you can create the new message of the day in the file /tmp/newmesg, and then issue the following command string:

at 2400 July 14 
chmod 000 /usr/public 
mv /etc/motd /etc/oldmotd 
mv /tmp/newmesg /etc/motd 
<Ctrl-D>

By default, any output of commands executed using at is mailed to the executing user through the system electronic mail. You can specify a different location for the disposition of output by using the standard output redirects, such as pipes (|) and file redirects (>). See your command shell documentation for a complete description of redirecting the standard output.

For complete information on the at command, see the at(1) reference page.

batch(1) Command

The batch command works just like the at command, except that you do not specify a time for the command or commands to be executed. The system determines when the overall load is low enough to execute the commands, and then does so. As with all other cron subsystem commands, the output of the commands is mailed to you unless you specify otherwise. batch is useful for large CPU-intensive jobs that slow down the system or cripple it during peak periods. If the job can wait until a non-peak time, you can place it on the batch queue until the system executes it. For complete information on the batch command, see the batch(1) reference page.

cron(1M) Command

If you desire to have a command executed regularly on schedule, the cron command and subsystem provide a precise mechanism for scheduled jobs. The at and batch commands are technically part of the cron subsystem and use cron to accomplish their tasks. The cron command itself, though, is the most configurable command of the subsystem.

You use cron by setting up a crontab file, where you list the commands you would like to have executed and the schedule for their execution. Complete information on setting up your crontab file is available in the cron(1M) and crontab(1) reference pages.

cron is useful for scheduling network backups, checking the integrity of the password file, and any other scheduled tasks that do not require interaction between you and the system. By default, cron mails the results or output of the command to the user who submitted the crontabs file, so if you use cron to schedule something like a pwck(1M), the results of the test are mailed to you and you can interpret them at your convenience.

Note that you must restart cron after each change to a crontabs file, whether made through the cron utility or the at command, for the changes to take effect.

Checking System Configuration

IRIX provides two commands that allow you to check your system hardware and software configurations. The hinv(1M) and versions(1M) commands display the hardware and software inventories, respectively.

hinv Command

The hinv command displays the machine's hardware inventory. hinv can be run from the Command Monitor or from your system shell prompt. Pertinent information such as the processor type, amount of main memory, and all disks, tape drives, or other devices is included. A sample hinv output for a typical workstation is:

1 100 MHZ IP22 Processor
FPU: MIPS R4010 Floating Point Chip Revision: 0.0
CPU: MIPS R4000 Processor Chip Revision: 3.0
On-board serial ports: 2
On-board bi-directional parallel port
Data cache size: 8 Kbytes
Instruction cache size: 8 Kbytes
Secondary unified instruction/data cache size: 1 Mbyte
Main memory size: 64 Mbytes
Vino video: unit 1, revision 1
Iris Audio Processor: version A2 revision 4.1.0
Integral Ethernet: ec0, version 1
CDROM: unit 4 on SCSI controller 0
Disk drive: unit 1 on SCSI controller 0
Integral SCSI controller 0: Version WD33C93B, revision D
Graphics board: Indy 24-bit

If a piece of peripheral hardware installed on your system does not appear in the hinv output, it may or may not be an indication of trouble with your hardware. Some peripherals connected to the system by a board on a VME bus will not be identified when running hinv from the Command Monitor. First, you should invoke hinv from a system shell prompt; If your peripheral is still not recognized, attempt to reseat the board in its socket and check that it is using the correct SCSI address. If this does not relieve the problem, the hardware itself may be defective. Note also that most devices are not recognized by hinv until the MAKEDEV(1M) command has been run after their installation.

versions Command

The versions command gives you an inventory of software packages that have been installed using inst(1M). This command can only be run at the system shell prompt, not from the Command Monitor. Software installed by other means is not included in the versions output. Along with the names of the software products, the release revision level numbers are displayed. By default, the output of versions includes all the products and their subsystems and is typically several hundred lines long, so it is often convenient to redirect the output to a file that you can view at your convenience. For a more general look at the products you have installed, without the list of specific subsystems, use the -b (brief) flag.

A sample versions -b output reads as follows (an actual listing will be much longer):

I = Installed, R = Removed
   Name        Date      Description
I  4Dwm        04/29/93  4Dwm -- Default Window Manager, 5.3
I  demos       04/29/93  Graphics Demonstration Program, 5.3
I  desktop_eoe 04/29/93  Desktop Environment, 5.3
I  dps_eoe     04/29/93  Display PostScript, 2.0 
I  eoe1        04/29/93  Execution Only Environment 1, 5.3
I  eoe2        04/29/93  Execution Only Environment 2, 5.3
I  insight     04/29/93  IRIS InSight Viewer, 2.1
I  motif_eoe   04/29/93  Motif Execution Only Environment
I  nfs         04/29/93  Network File System, 5.2

gfxinfo command

The gfxinfo command is useful for determining the graphics hardware installed in the system. It is in the /usr/gfx directory, which is not on any of the standard search paths. Thus gfxinfo typically needs full path name for execution. It requires no arguments to run.

An sample output for an Indy:

% /usr/gfx/gfxinfo 
Graphics board 0 is "NG1" graphics.
        Managed (":0.0") 1280x1024 
        24 bitplanes, NG1 revision 3, REX3 revision B,
        VC2 revision A
        MC revision C, xmap9 revision A, cmap revision C,
        bt445 revision A
        Display 1280x1024 @ 60Hz, monitor id 12

This command provides much more information about the graphics system than the hinv command (hinv would simply return Indy 24-bit). From the output of gfxinfo you can determine the number of screens and their pixel resolutions, bitplane configurations, component revision levels, and monitor types. There is no reference page for gfxinfo.

uname command

uname returns information such as the OS version and hostname. The -a options gives a complete list of the uname output. See the reference page for a description of all the uname options and fields.

lpstat command

lpstat with the -a option will show all the printers configured for the lp spooling system and also give their status. For more information on the lpstat command, see "lpstat: Report lp Status" or the lpstat reference page.

General Operations

This section covers general system operations. Topics are:

Information on specific parts of the system is presented in other chapters. For example, information on system security is covered in Chapter 12, "System Security."

Also, defaults for tape and other backup operations (such as the default tape drive) are covered in Chapter 6, "Backing Up and Restoring Files." Defaults for shell types and user groups are covered in Chapter 3, "User Services."

Checking the System Configuration

You can quickly check the configuration of a workstation or server with the chkconfig(1) command. The /sbin/chkconfig utility reports the state of various process daemons (that is, whether or not they are supposed to be active).

For example, enter the chkconfig command:

chkconfig

You see a display similar to this:

Flag                 State
====                 =====
acct                 off
audit                off
automount            on
fmlicserv            off
gated                off
lockd                on
mrouted              off
named                off
network              on
nfs                  on
noiconlogin          off
nsr                  on
quotacheck           off
quotas               off
routed               on
rtnetd               off
rwhod                off
sar                  on
snmpd                on
timed                on
timeslave            off
verbose              off
visuallogin          on
windowsystem         off
yp                   on
ypmaster             off
ypserv               off

This example is typical for a networked workstation with the Network File System (NFS) option installed. The left column of the output describes a system feature, and the right column indicates whether it is on or off. The following list provides more specific information about each system feature:

acct

Detailed system accounting is turned on or off.

audit

The System Audit Trail is turned on or off.

automount

The NFS automount(1M) daemon is turned on or off. This configuration option is available only if you have NFS installed on the workstation.

gated

The gated(1M) daemon, which manages multiple routing protocols is turned on or off.

glb

This option is used by the NetLS license server for the global location broker daemon.

llb

This option is used by the NetLS license server for the local location broker daemon.

lockd

The Network File System (NFS) lock daemon is turned on or off. This configuration option is available only if you have NFS installed on the workstation.

mrouted

The Stanford IP multicast routing daemon is turned on or off.

named

named(1M), the Internet domain name server, is turned on or off.

network

The network is turned on or off.

nfs

NFS is turned on or off. This configuration option is available only if you have NFS installed on the workstation.

noiconlogin

The visual login program, pandora(1), displays icons that represent users on the system. This feature does not enable or disable pandora; it only affects whether or not pandora displays icons. It is turned on or off. To enable or disable pandora, use the visuallogin feature.

nsr

IRIS Networker backup utility. This configuration option is available only if you have Networker installed on the workstation.

quotacheck

The disk space quota checker is enabled or disabled.

quotas

Disk quotas are enabled or disabled.

routed

routed(1M), which manages the network routing tables, is turned on or off.

rtnetd

rtnetd(1M), which allows higher priority real-time processes to preempt processing of incoming network packets, is turned on or off.

rwhod

rwhod(1M) is turned on or off.

sar

sar(1), the system activity reporter, is turned on or off.

snmpd

The Simple Network Management Protocol Daemon is turned on or off.

timed

timed(1M), the 4.3 BSD time server daemon, is turned on or off.

timeslave

The Silicon Graphics time server daemon is turned on or off. Like timed, this attaches a workstation's clock to a different clock, usually some kind of master time server for a group of workstations or for the entire site.

verbose

If this feature is enabled, as the system boots or is shut down, daemons print information about their functions. If this feature is disabled, less information is printed when the system is started and shut down.

visuallogin

The visual login program, pandora(1), is turned on or off.

windowsystem


The window manager is turned on or off.

yp

The network information service (NIS) is enabled on or off. This is called "yp" for historical reasons. NIS is available with the NFS software. This configuration option is available only if you have NFS installed on the workstation.

ypmaster

NIS master services are turned on or off. This configuration option is available only if you have NFS installed on the workstation.

ypserv

NIS server and bind processes are turned on or off. This configuration option is available only if you have NFS installed on the workstation.

Note that if a daemon is enabled using chkconfig, it does not necessarily mean that the daemon starts up immediately, or that it is running successfully. To verify that a daemon is running, use the ps(1) command to identify what processes are running on the system. For example, the command:

ps -ef

produces output similar to this:

Table 2-2 : ps -ef Output
UID PID PPID C STIME TTY TIME COMMAND
root 0 0 0 Aug 3 ? 0:00 sched
root 1 0 0 Aug 3 ? 0:45 /etc/init
root 2 0 0 Aug 3 ? 0:08 vhand
root 3 0 0 Aug 3 ? 0:09 bdflush



This example is edited for simplicity. An actual, full ps listing shows many more active processes.

To view information about specific processes, and avoid searching through a large ps listing, you can filter the listing with the grep(1) or egrep(1) commands. For example, to look at process information for only the NFS daemons, use this command:

ps -ef | egrep 'nfsd|biod'

The output of this command is similar to this (assuming you have NFS installed and running):

root  120 1  0    09:40:05 ?     0:02 /usr/etc/nfsd 4 
root  122 12 0 0  09:40:05 ?     0:02 /usr/etc/nfsd 4 
root  123 12 0 0  09:40:05 ?     0:02 /usr/etc/nfsd 4 
root  124 12 0 0  09:40:05 ?     0:02 /usr/etc/nfsd 4 
root  126 1  0    09:40:05 ?     0:00 /usr/etc/biod 4 
root  127 1  0    09:40:05 ?     0:00 /usr/etc/biod 4 
root  128 1  0    09:40:05 ?     0:00 /usr/etc/biod 4 
root  129 1  0    09:40:05 ?     0:00 /usr/etc/biod 4 
root  131 1  0    09:40:11 ?     0:00 /etc/mount -at nfs 
ralph 589 55 0 11 15:25:30 ttyq1 0:00 egrep nfsd|biod

Note that the final entry in the ps listing is the process that produced the listing and that it is the only non-root process to have nfsd or biod in its name.

Altering the System Configuration

You can use the chkconfig command to change some aspects of system configuration. To determine which aspects of a system you can alter with

chkconfig, enter the chkconfig command:

chkconfig

You see a list of configuration options, which are described in "Checking the System Configuration" If you use the -s option, you see a list that is sorted by whether the configuration item is on or off.

To change a configuration option, use the chkconfig command with two arguments: the name of the option you wish to change and the new status of the configuration (on or off). You must have root privilege to change a system configuration.

For example, to turn on detailed process accounting, log in either as root or as the system administrator, and enter:

chkconfig acct on

To turn off process accounting, enter:

chkconfig acct off

Some aspects of system configuration do not take effect until the system is shut down and rebooted because startup scripts, which are in the directory /etc/init.d, are run when the system is booted up and brought to multiuser mode. These scripts read the files that chkconfig sets to determine which daemons to start.

Some configuration items that can be controlled by chkconfig may not be displayed by chkconfig. These include:

nostickytmp

Sets ``sticky'' behavior for the directory /tmp. When the directory is sticky, (with nostickytmp set to off), users may not remove files from the directory unless they own the files, have explicit permission to remove the files (write permission), or have superuser privileges.

The opposite behavior allows users to remove or replace files in /tmp, which is a publicly writable directory, even if they do not own the files. This is handy behavior if you have users who need to create large temporary files and you are short on disk space. But it is better to increase disk space to avoid important files being removed.

nocleantmp

Controls whether or not the directory /tmp is cleaned out each time the system is booted. If nocleantmp is on, /tmp is not cleaned. If nocleantmp is off, all files in /tmp are removed each time the system is started.

If you want to see these flags in the chkconfig menu, you can use the -f option to force chkconfig to create a configuration file for the options:

chkconfig -f nocleantmp on

In this example, chkconfig creates a configuration file called nocleantmp in the directory /etc/config.

Checking the Password File

At least once a week, you should run the pwck(1M) and grpck(1M) programs to check your /etc/passwd and /etc/group files for errors. You can automate this process using the cron(1) command, and you can direct cron to mail the results of the checks to your user account. For more information on using cron to automate your routine tasks, see "Automating Tasks with at(1), batch(1), and cron(1M)".

The pwck and grpck commands read the password and group files and report any incorrect or inconsistent entries. Any inconsistency with normal IRIX operation is reported. For example, if you have /etc/passwd entries for two user names with the same User Identification (UID) number, pwck will report this as an error. grpck performs a similar function on the /etc/group file. Note that the standard passwd file shipped with the system generates several errors.

Changing System Defaults

These system-wide defaults affect programs and system functions:

Some of these defaults are described more thoroughly in specific sections of this guide, but they are all presented here to provide a more thorough overview of the IRIX system.

Setting the System Display

You can make the output of programs and utilities running on one system appear on the screen of another system on the same network by changing the DISPLAY environment variable. This is useful if your network includes graphical systems and non-graphical servers. In order to view information from the server graphically, you reset the display to a graphics workstation.

For example, if your server has only a character-based terminal as its console and you wish to run gr_osview(1M) to visually inspect your CPU usage, you would issue commands similar to these on the server:

setenv DISPLAY graphics_machine:0 
gr_osview

When you invoke gr_osview, the window with the output will appear on the machine name you specify. In this example, graphics_machine was used in place of the system name. The :0 used after the machine name indicates that display monitor 0 (the graphics console) should be used to display the output. When you have finished using the graphics console, be sure to reset the display by issuing this command on the server:

setenv DISPLAY local_server:0

where local_server is the name of your server.

Setting the Time Zone

To set the time zone of the system, edit the file /etc/TIMEZONE. For a site on the east coast of the United States, the file might look something like this:

# Time Zone
TZ=EST5EDT

The line TZ=EST5EDT means:

The TZ environment variable is read by init(1) when the system boots, and the value of TZ is passed to all subsequent processes. For complete information about setting your time zone, see the timezone(4) reference page.

Changing Processors on Multi-Processor Systems

If you have a multi-processor system, the mpadmin(1M) and pset(1M) commands allow you to change the way programs are assigned to the various processors on your system. To determine if your system is multi-processor, use the hinv(1M) command. A multi-processor system returns information similar to the following in its hinv output:

Processor 0: 40 MHZ IP7 
Processor 1: 36 MHZ IP7 
Processor 2: 40 MHZ IP7 
Processor 3: 40 MHZ IP7 
Processor 4: 40 MHZ IP7 
Processor 5: 40 MHZ IP7 
Processor 6: 40 MHZ IP7 
Processor 7: 40 MHZ IP7

Or, alternately, output similar to the following:

8 40 MHZ IP7 Processors

A single-processor system returns information similar to the following for the hinv command:

1 100 MHZ IP22 Processor

If you have only one processor on your system (and the vast majority of systems have only one processor) these commands still operate, though they have no useful purpose.

The mpadmin command allows you to ''turn off'' processors, report various states of the processors, and move system functions such as the system clock to specific processors. The pset command is used both to display and modify information concerning the use of processor sets and programs running in the current system. The pset command provides a much more detailed level of control of processes and processors.

For complete information on mpadmin(1M) and pset(1M), see the respective reference pages.

Changing the Name of a System

The name of the system is stored in several places. If you wish to change the name of your system, you must change all these files together or your system will not function correctly:

Note that you should not arbitrarily change the name of a running workstation. Many programs that are started at boot time depend on the name of the workstation.

To display the name of the system, use the hostname command with no arguments:

hostname

This displays the name of the system. The uname command also displays the name of the system, along with other information.

To change the name of the workstation, follow these steps:

    Log in as root.

    Edit the file /etc/sys_id. Change the name of the host to newname. Write and exit the editor.

    You must also change the name of the host in any network files, such as /etc/hosts, and possibly in the NIS map on the master NIS server.

    Reboot your system.

The name of the workstation is now changed. When the workstation is booted, all programs that are started at boot time, and read the host name when they start, now use the correct host name.

For information about the Internet address of a workstation, see Chapter 15, "Understanding Silicon Graphics' Networking Products." For more information about the name of the system, see the hostname(1) and uname(1) reference pages.

Setting the Network Address

The system's network address (IP address) is covered more thoroughly in Chapter 15, "Understanding Silicon Graphics' Networking Products."

To set the network address, follow these steps:

    Place the network address in /etc/hosts on the same line as the system name.

    If you use the network information service (NIS), place the name of your domain in the file /var/yp/ypdomain, if it is installed.

    Use the nvram(1M) command to set the variable netaddr to the IP number of the machine. For example:

    nvram netaddr 192.13.52.4

Setting the Default Printer

The lpadmin(1M) command sets the default printer. This command sets the default printer to laser:

lpadmin -dlaser

Note that the printer laser must already exist and be configured. For complete information on setting up printers, see Chapter 9, "Administering Printers."

Managing User Processes

Just as files can use up your available disk space, too many processes going at once can use up your available CPU time. When this happens, your system response time gets slower and slower until finally the system cannot execute any processes effectively. If you have not tuned your kernel to allow for more processes, the system will refuse new processes long before it reaches a saturation point. However, due to normal variations in system usage, you may experience fluctuations in your system performance without reaching the maximum number of processes allowed by your system.

Monitoring User Processes

Not all processes require the same amount of system resources. Some processes, such as database applications working with large files, tend to be disk intensive, requiring a great deal of reading from and writing to the disk as well as a large amount of space on the disk. These activities take up CPU time. Time is also spent waiting for the hardware to perform the requested operations. Other jobs, such as compiling programs or processing large amounts of data, are CPU intensive, since they require a great number of CPU instructions to be performed. Some jobs are memory intensive, such as a process that reads a great deal of data and manipulates it in memory. Since the disk, CPU, and memory resources are limited, if you have more than a few intensive processes running at once on your system, you may see a performance degradation.

As the administrator, you should be on the lookout for general trends in system usage, so you can respond to them and keep the systems running as efficiently as possible. If a system shows signs of being overloaded, and yet the total number of processes is low, your system may still be at or above reasonable capacity. The following sections show four ways to monitor your system processes.

Monitoring Processes with top

top and gr_top are the most convenient utilities provided with IRIX to monitor the top CPU-using processes on your system. These utilities display the top such processes dynamically, that is, if a listed process exits, it is removed from the table and the next-highest CPU-using process takes its place. gr_top graphically displays the same information as top. If you are using a non-graphics server, you cannot use gr_top locally, but you can use it if you set the display to another system on the network that does have graphics capability. For complete information on configuring and using top and gr_top, consult the top(1) and gr_top(1) reference pages. For information on resetting the display, see "Setting the System Display".

Monitoring Processes with osview

osview and gr_osview are utilities that display kernel execution statistics dynamically. If you have a graphics workstation, you can use the gr_osview(1) tool, which provides a real-time graphical display of system memory and CPU usage. osview provides the same information in ASCII format. You can configure gr_osview to display several different types of information about your system's current status. In its default configuration, gr_osview provides information on the amount of CPU time spent on user process execution, system overhead tasks, interrupts, and idle time. For complete information on osview and gr_osview, see the osview(1) and gr_osview(1) reference pages.

Monitoring Processes with sar

The System Activity Reporter, sar, provides essentially the same information as osview, but it represents a ``snapshot'' of the system status, not a dynamic reflection. Because sar generates a single snapshot, it is easily saved and can be compared with a similar snapshot taken at another time. You can use sar automatically with cron to get a series of system snapshots over time to help you locate chronic system bottlenecks. For complete information on sar, see the sar(1) reference page.

Monitoring Processes with ps

The ps - ef command allows you to look at all the processes currently running on your system. ps - ef output looks very similar to Table 2-3:

Table 2-3 : Output of the ps -ef Command
Name PID PPID C Time TTY CPU Time Process
joe 23328 316 1 May 5 ttyq1 1:01 csh



In this table, the process shown is for the user ``joe.'' In a real situation, each user with processes running on the system is represented. Each field in the output contains some useful information.

Name

The login name of the user who ''owns'' the process.

PID

The process identification number.

PPID

The process identification number of the parent process that spawned or forked the listed process.

C

Current execution priority. The higher this number, the lower the scheduling priority. This number is based on the recent scheduling of the process and is not a definitive indicator of its overall priority.

Time

The time when the process began executing. If it began more than 24 hours before the ps command was given, the date on which it began is displayed.

TTY

The TTY (Terminal or window) with which the process is associated.

CPU

The total amount of CPU time expended to date on this process. This field is useful in determining which processes are using the most CPU time. If a process uses a great deal in a brief period, it can cause a general system slowdown.

For even more information, including the general system priority of each process, use the -l flag to ps. For complete information on interpreting ps output, see the ps(1) reference page.

Prioritizing Processes with nice

IRIX provides methods for users to force their CPU-intensive processes to execute at a lower priority than general user processes. The /bin/nice(1) and npri(1M) commands allow the user to control the priority of their processes on the system. The nice command functions as follows:

nice [ -


increment ] 


command

When you form your command line using /bin/nice, you fill in the increment field with a number between 1 and 19. If you do not fill in a number, a default of 10 is assumed. The higher the number you use for the increment, the lower your process' priority will be (19 is the lowest possible priority; all numbers greater than 19 are interpreted as 19). The csh(1) shell has its own internal nice functions, which operate differently from the nice command, and are documented in the csh(1) reference page.

After entering the nice command and the increment on your command line, give the command as you would ordinarily enter it. For example, if the user ``joe'' wants to make his costly compile command described in the ps -ef listing above happen at the lowest possible priority, he forms the command line as follows: 

nice -19 cc -o prog prog.c

If a process is invoked using nice, the total amount of CPU time required to execute the program does not change, but the time is spread out, since the process executes less often.

The superuser (root) is the only user who can give nice a negative value and thereby increase the priority of a process. To give nice a negative value, use two minus signs before the increment. For example: 

nice --19 cc -o prog prog.c

The above command endows that process with the highest priority a user process may possess. The superuser should not use this feature frequently, as even a single process that has been upgraded in priority causes a significant system slowdown for all other users. Note that /bin/csh has a built-in nice program that uses slightly different syntax than that described here. For complete information on csh, see the csh(1) reference page.

The npri command allows users to make their process' priority nondegrading. In the normal flow of operations, a process loses priority as it executes, so large jobs typically use fewer CPU cycles per minute as they grow older. (There is a minimum priority, too. This priority degradation simply serves to maintain performance for simple tasks.) By using npri, the user can set the nice value of a process, make that process non-degrading, and also set the default time slice that the CPU allocates to that process. npri also allows you to change the priority of a currently running process. The following example usage of npri sets all the possible variables for a command: 

npri -h 10 -n 10 -t 3 cc -o prog prog.c

In this example, the -h flag sets the nondegrading priority of the process, while the -n flag sets the absolute nice priority. The -t flag sets the time slice allocated to the process. IRIX uses a 10-millisecond time slice as the default, so the example above sets the time slice to 30 milliseconds. For complete information about npri and its flags and options, see the npri(1) reference page.

Changing the Priority of a Running Process

The superuser can change the priority of a running process with the renice(1M) or npri commands. Only the superuser can use these commands. renice is used as follows: 

renice 


increment pid [-u 


user] [-g 


pgrp]

In the most commonly used form, renice is invoked on a specific process that is using system time at an overwhelming rate. However, you can also invoke it with the -u flag to lower the priority of all processes associated with a certain user, or with the -g flag to lower the priorities of all processes associated with a process group. More options exist and are documented in the renice(1M) reference page.

The npri command can also be used to change the parameters of a running process. This example changes the parameters of a running process with npri:

npri -h 10 -n 10 -t 3 -p 11962

The superuser can use renice or npri to increase the priority of a process or user, but this can severely impact system performance.

Terminating Processes

From time to time a process may use so much memory, disk, or CPU time that your only alternative is to terminate it before it causes a system crash. Before you kill a process, make sure that the user who invoked the process will not try to invoke it again. You should, if at all possible, speak to the user before killing the process, and at a minimum you should notify the user that the process was prematurely terminated and give a reason for the termination. If you do this, the user can reinvoke the process at a lower priority or possibly use the system's job processing facilities (at, batch, and cron) to execute the process at another time.

To terminate a process, you use the kill command. Typically, for most terminations, you should use the kill -15 variation. The -15 flag indicates that the process is to be allowed time to exit gracefully, closing any open files and descriptors. The -9 flag to kill terminates the process immediately, with no provision for cleanup. If the process you are going to kill has any child processes executing, using the kill -9 command may cause those child processes to continue to exist on the process table, though they will not be responsive to input. The wait(1) command, given with the process number of the child process, removes them. For complete information about the syntax and usage of the kill command, see the kill(1) reference page. You must always know the PID of the process you intend to kill with the kill command.

Killing Processes by Name with the killall(1M) Command

The killall(1M) command allows you to kill processes by their command name. For example, if you wish to kill the program a.out that you invoked, use the syntax:

killall a.out

This command allows you to kill processes without the time-consuming task of looking up the process ID number with the ps(1M) command.

Note: This command kills all instances of the named program running under your shell and if invoked with no arguments, kills all processes on the system that are killable by the user who invoked the command. For ordinary users, these are simply the processes invoked and forked by that user, but if invoked by root, all processes on the system will be killed. For this reason, this command should be used carefully.

Changing the Date and Time

Use the date(1) command to set the date and time. For example, to set the date to April 1st, 1999, and the time to 09:00, log in as root and enter: 

date 0401090099

Changing the date and time on a running system can have unexpected consequences. Users and administrators use system scheduling utilities (at(1), cron(1), and batch(1)) to perform commands at specified times. If you change the effective date or time on the system, these commands may not execute at the desired times. Similarly, if your users use the make(1) utility provided with the system, the commands specified in Makefiles can incorrectly compile or process your users' work. Always try to keep your system date and time accurate within reason. Random changes of the date and time can be extremely inconvenient and possibly destructive to users' work.

If timed(1M) is running on the system, and it is a slave system, the time is reset by timed and not the above command. For more information, see the timed(1M) reference page.

Creating a Message of the Day

The file /etc/motd contains the ``message of the day.'' This message is displayed on a user's screen, either by /etc/profile if the user runs Bourne shell, or by /etc/cshrc if the user runs C shell, when the user first logs in to the system.

You can place announcements of system activity in the motd file. For example, you should warn users of scheduled maintenance, changes in billing rates, new hardware and software, and any changes in the system or site configuration that affect them.

Since users see this message every time they log in, you should change it frequently to keep it from becoming stale. If users see the same message repeatedly, they lose interest in reading the message of the day and can miss an important announcement.

Make sure you remove outdated announcements. If nothing new is happening on the system, trim the file to a short ``welcome to the system'' message.

A typical motd file looks something like this:

Upcoming Events: ---------------

26 November -- The system will be down from 8PM until Midnight for a software upgrade. We are installing FareSaver+, Release 3.2.2d.

Watch this space for further details.

28 November through 31 November -- We will be operating with a minimal staff during the holiday. Please be patient if you need computer services. Use the admin beeper (555-3465) if there is a serious problem.

The motd file is used more frequently on servers than on workstations, but can be handy for networked workstations with guest accounts. You can also send electronic mail to all people who use the system, but this method consumes more disk space, and users may accidentally skip over the mail in their mailboxes.

Creating a Remote Login Message

The file /etc/issue is displayed before the login prompt is given to someone attempting to log in to a system over a serial line or the network. If /etc/issue does not exist, or is empty, no message is displayed. This is essentially different from /etc/motd because it is displayed before the login prompt. This message is often used to notify potential users of rules about using the equipment; for example, a disclaimer that the workstation or server is reserved for employees of a particular corporation and that intruders will be prosecuted for unauthorized use.

Maintaining the File Alteration Monitor

The File Alteration Monitor (famd) is a daemon that monitors files and directories. Application writers can include certain function calls in their applications to let famd know that they want to be informed of changes to whichever files and/or directories they specify. WorkSpace and Mailbox are two applications that use famd; WorkSpace uses it to keep the directory views up to date, and Mailbox uses it to know when to indicate the arrival of new mail.

The famd daemon runs only when applications using it are running; it exits after all programs using it have stopped.

Sometimes, when attempting to start up an application which uses famd, an error message is displayed:

Cannot connect with File Alteration Monitor (fam)

There are several reasons why this message appears. Below are some of the common ways to troubleshoot the problem.

Using a Foreign NIS Master

If you have the optional NIS (YP) software installed at your site, and you are using another manufacturer's system as your NIS master, with no rpc entries for sgi_toolkitbus and sgi_fam, this section provides the information to correct the error message.

Depending on the operating system (the Sun 3.x or the Sun 4.0) on the Sun NIS (YP) server, one of the two following solutions applies.

Verify fam Installation

From a shell window, enter the command: 

ls -l /usr/etc/fam

If the file is not found, you must reinstall the software package:

eoe2.sw.envm.

Check Network Activity

The famd daemon is enabled via the /usr/etc/inetd.conf file, which is read when the system starts up the network. Even if the system is not connected via the network to other systems, the network software must be started.

If the message:

portmapper failure

is displayed, it is also a sign that the network is not active.

If the command:

/etc/chkconfig | grep network

returns the message:

network off

Turn the network on by entering the command (as root):

/etc/chkconfig network on

and then either reboot your system by entering the command:

reboot

or just start up the network by issuing the command:

/etc/init.d/network start

"localhost" Entry in /etc/hosts

Applications connect to famd via the localhost. Therefore, there must be a localhost entry in the /etc/hosts file.

This entry is in the hosts file by default, but sometimes gets removed. The correct entry would be:

127.0.0.1 localhost

sgi_famd/sgi_toolkitbus

Make sure that both sgi_famd and sgi_toolkitbus are running by entering the command:

/usr/etc/rpcinfo -p

In addition to other daemons running, you should see the following entries:

391002 1 tcp 1086 sgi_fam 
391001 1 tcp 1087 sgi_toolkitbus

If you do not see these lines, check your /usr/etc/inetd.conf file. The following entries must be in that file:

sgi_fam/1 stream rpc/tcp wait root /usr/etc/fam famd
sgi_toolkitbus/1 stream rpc/tcp wait root /usr/etc/rpc.toolkitbus toolkitbusd

If those lines are not in the file, you must add them, and then restart your network by entering the following command, as root:

/etc/init.d/network stop 
/etc/init.d/network start

vadmin Permissions

The permissions on the /usr/lib/vadmin file should be:

drwxr-xr-x 2 root sys

eoe2.sw.fonts

If none of the listed problems exists, reinstall eoe2.sw.fonts. This will install a file called /usr/sysgen/boot/imon.o. Confirm that the file is there, and then use the autoconfig-vf command to build a new kernel. Then reboot your system to install the new kernel.

Operating Policies

This section describes general operating policies for you, the site administrator. For a discussion of site policies, especially as they affect your users, see "System and Site Policies" in this guide.

Shutting Down the System Carefully

Many administrative tasks require the system to be shut down to a run level other than the multiuser state. This means that conventional users cannot access the system. Just before the system is taken out of the multiuser state, users on the system are requested to log off. You should do these types of tasks when they will interfere the least with the activities of the user community.

Sometimes situations arise that require the system to be taken down with little or no notice provided to the users. This is often unavoidable, but try to give at least five to fifteen minutes of notice, if possible.

In general, try to provide the user community as much notice as possible about events affecting the use of the system. When the system must be taken out of service, also tell the users when to expect the system to be available. Use the ``message of the day'' file /etc/motd to keep users informed about changes in hardware, software, policies, and procedures.

At your discretion, the following actions should be prerequisites for any task that requires the system to leave the multiuser state:

Maintaining a System Log Book

It is important to keep a complete set of records about each system you administer. A system log book is a useful tool when troubleshooting transient problems or when trying to establish system operating characteristics over a period of time. Keeping a hardcopy book is important, since you won't be able to refer to an online log if you have trouble starting the system.

Some of the things that you should consider entering into the log book are:

The format of the system log and the types of items noted in the log should follow a logical structure. Think of the log as a diary that you update periodically. To a large measure, how you use your system will dictate the form and importance of maintaining a system log.

Administrative Directories and Files

This section briefly describes the directories and files that a system administrator uses frequently. For additional information on the formats of the system files, refer to the IRIX reference pages.

Appendix B, "IRIX Device Files,"contains further information on the device files and directories that reside in the /dev directory.

Root Directories

The main directories of the root file system (/) are as follows:

/

Contains hardware-specific files and files required to start the system.

bin

Contains publicly executable commands. (Some are root-only.)

debug

Provides a link to /proc.

dev

Contains special files that define all of the devices on the system.

etc

Contains administrative programs and tables.

lib

Contains public libraries.

lost+found

Used by fsck(1M) to save disconnected files and directories.

proc

Provides an interface to running processes that may be used by debuggers such as dbx(1).

tmp

Used for temporary files.

usr

Used to mount the /usr file system and for files that are the same from system to system. These files are not writable.

var

Used for files that are specific to each system. There is typically a symbolic link to /usr for each file in /var.

Important System Directories

The following directories are important in the administration of your system:

/etc/init.d

Contains shell scripts used in upward and downward transitions to all system run levels. These files are linked to files beginning with S (start) or K (kill) in /etc/rcn.d, where n is replaced by the appropriate run level number.

/etc/config

Contains start-up and run-time configuration information.

/etc/rc0.d

Contains files executed by /etc/rc0 to bring the system to run-level 0. Files in this directory are linked from files in the /etc/init.d directory and begin with either a K or an S. K indicates processes that are killed, and S indicates processes that are started when entering run-level 0.

/etc/rc2.d

Contains files executed by /etc/rc2 for transitions to system run-level 2. Files in this directory are linked from files in the /etc/init.d directory and begin with either a K or an S. K indicates processes that should be killed, and S indicates processes that should be started, when entering run-level 2.

/etc/rc3.d

Contains files executed by /etc/rc3 for transitions to system run-level 3. Files in this directory are linked from files in the /etc/init.d directory and begin with either a K or an S. K indicates processes that should be stopped, and S indicates processes that should be started when entering run-level 3.

/var/adm/acct


Contains information collected by the accounting subsystem.

/var/adm/crash


Contains crash dumps of the system. After analysis, and if appropriate, these dumps can safely be removed unless your support provider has requested otherwise. See the savecore(1) reference page for more information.

/var/adm/sa

Contains information collected by sar(1).

/usr/people

Contains the home directories of users of the system or network. This directory can be a link to /var/people or a mount point for a totally separate file system.

/usr/share

This directory contains files that are the same on all systems.

/var/spool

Contains spooling directories. The directories in this directory hold outbound mail, print requests, and other data.

/var/spool/cron/crontabs


Contains crontab files for the adm, root, and sys logins and ordinary users listed in cron.allow.

/var/sysgen/master.d


Contains files that define the configuration of hardware devices, software services and utilities, and aliases.

/var/sysgen/stune


Contains files that define the default settings of all kernel tunable parameters.

/var/sysgen/mtune


Contains files that define the current settings of all kernel tunable parameters.

Important System Files

The following files are important in the administration of your system:

/etc/cshrc

Contains the standard (default) environment for /bin/csh users.

/etc/exports

Contains the list of NFS file systems exported at boot time to NFS clients if the optional NFS software is installed.

/etc/fstab

Specifies the file system(s) to be mounted by /etc/mountall if the optional NFS software is installed.

/etc/gettydefs

Contains information used by getty to set the speed and terminal settings for a line.

/etc/group

Describes each group to the system.

/etc/hosts

Contains information about the known hosts on the network.

/etc/hosts.equiv


Contains a list of hosts trusted for non-superuser rlogin and rsh execution.

/etc/inittab

Contains the instructions to define the processes created or terminated by init for each initialization state.

/etc/issue

Displays a message to users before logging in to the system over the network or on serial lines.

/etc/lvtab

Contains information describing the logical volumes used by the workstation. This file is read by the logical volumes utilities.

/etc/motd

Contains a brief ``message of the day.''

/etc/passwd

Identifies each user to the system.

/etc/profile

Contains the standard (default) environment for /bin/sh users.

/etc/rc0

Contains a script that executes shell scripts in /etc/rc0.d to bring the system to run-level 0.

/etc/rc2

Contains a script that executes shell scripts in /etc/rc2.d and /etc/rc.d on transitions to system run-level 2.

/etc/shutdown


Contains a shell script that gracefully shuts down the system in preparation for system backup or for scheduled downtime.

/etc/sys_id

Contains the system name.

/etc/ttytype

Contains a list, ordered by terminal port, of what kind of terminal is likely to log in to that port.

/etc/TIMEZONE


Used to set the default time zone shell variable TZ.

/etc/utmp

Contains the information on the current runstate of the system.

/etc/wtmp

Contains a history of system logins.

/etc/xwtmp

Contains an extended history of system logins.

/var/adm/sulog


Contains a history of su command usage. This file should be checked periodically for excessive size and archived.

/var/adm/SYSLOG


Contains system and daemon error messages.

/var/yp/ypdomain


Contains the domain name if the workstation is using NIS.

/var/cron/log


Contains a history of all the actions taken by cron. This file should be checked periodically for excessive size and reduced if necessary.

/usr/lib/cron/cron.allow


Contains a list of users allowed to use crontab(1). This file cannot exist on the system at the same time as cron.deny.

/usr/lib/cron/cron.deny


Contains a list of users who are denied access to crontab(1). It is checked if /usr/lib/cron/cron.allow does not exist.

Operating Levels

The IRIX system can run in either single-user or multiuser mode. In single-user mode, only a few processes are active on the system, no graphics are available, and only a single login is allowed. In multiuser mode, there can be multiple login sessions, many files open at once, and many active processes, including numerous background daemons.

The init program controls whether the system is in the multiuser or single-user state. Each possible state that the system can be in is assigned a label, either a number or a letter. The shutdown state is state 0. Single-user mode is state s.

Multiuse state labeling is more complex, because there can be great variations in multiuser states. For example, in one multiuser state, there can be unlimited logins, but in another state there can be a restricted number of logins. The different states can all be assigned different numbers.

The state of the system is controlled by the file /etc/inittab. This file lists the possible states, and the label associated with each.

When you bring the system to standard multiuser mode, init state 2, the following happen:

Not all activities can or should be performed in the multiuser state. Some tasks, such as installing software that requires the miniroot and checking file systems must be done with the system in single-user mode.

There are many synonyms for the system state. These include:

Likewise, each system state may be referred to in a number of ways; for example:

Table 2-4 shows the various possible states of the operating system as it is shipped. You can, of course, create your own custom states.

Table 2-4 : System States
Run Level Description
0 Power-down state.
1, s, or S Single-user mode is used to install/remove software utilities, run file system backups/restores, and check file systems. This state unmounts everything except root, and kills all user processes except those that relate to the console.
2 Multiuser mode is the normal operating mode for the system. The default is that the root (/) and user (/usr) file systems are mounted in this mode. When the system is powered up, it is put in multiuser mode.
6 Reboot mode is used to bring down the system and then bring it back up again. This mode is useful when you are changing certain system configuration parameters.



How init Controls the System State

The init process is the first general process created by the system at startup. It reads the file /etc/inittab, which defines exactly which processes exist for each run level.

In the multiuser state (run level 2), init scans the file for entries that have a tag (also 2) for the run level and executes everything after the third colon on the line containing the tag. For complete information, see the inittab(4) reference page.

The system /etc/inittab looks something like this:

is:2:initdefault: 
fs::sysinit:/etc/bcheckrc </dev/console >/dev/console 2>&1 
mt::sysinit:/etc/brc </dev/console >/dev/console 2>&1 
s0:06s:wait:/etc/rc0 >/dev/console 2>&1 </dev/console
s1:1:wait:/etc/shutdown -y -iS -g0 >/dev/console 2>&1 </dev/console 
s2:23:wait:/etc/rc2 >/dev/console 2>&1 </dev/console 
s3:3:wait:/etc/rc3 >/dev/console 2>&1 </dev/console 
or:06:wait:/etc/umount -ak -b /proc,/debug > /dev/console 2>&1
of:0:wait:/etc/uadmin 2 0 >/dev/console 2>&1 </dev/console 
RB:6:wait:echo "The system is being restarted." >/dev/console 
rb:6:wait:/etc/uadmin 2 1 >/dev/console 2>&1 </dev/console 
#

This display has been edited for brevity; the actual /etc/inittab file is lengthy. If /etc/inittab is removed by mistake and is missing during shutdown, init enters the single-user state (init s). While entering single-user state, /usr remains mounted, and processes not spawned by init continue to run. Immediately replace /etc/inittab before changing states again. The format of each line in inittab is:

id:level:action:process

where:

When changing levels, init kills all processes not specified for that level.

Entering the Multiuser State from System Shutdown

When your system is up and running, it is usually in multiuser mode. It is only in this mode that the full power of IRIX is available to your users.

Powering Up the System

When you power up your system, it enters multiuser mode by default. (You can change the default by modifying the initdefault line in your inittab file.) In effect, going to the multiuser state follows these stages (see Table 2-4, "System States,"):

    The operating system loads and the early system initializations are started by init.

    The run level change is prepared by the /etc/rc2 procedure.

    The system is made public through the spawning of getty processes along the enabled terminal lines.

Early Initialization

Just after the operating system is first loaded into physical memory through the specialized boot programs that are resident in the PROM hardware, the init process is created. It immediately scans /etc/inittab for entries of the type sysinit:

fs::sysinit:/etc/bcheckrc </dev/console >/dev/console 2>&1 
mt::sysinit:/etc/brc </dev/console >/dev/console 2>&1

These entries are executed in sequence and perform the necessary early initializations of the system. Note that each entry indicates a standard input/output relationship with /dev/console. This establishes communication with the system console before the system is brought to the multiuser state.

Preparing the Run Level Change

Now the system is placed in a particular run level. First, init scans the table to find an entry that specifies an action of the type initdefault. If it finds one, it uses the run level of that entry as the tag to select the next entries to be executed. In our sample /etc/inittab, the initdefault entry is run level 2 (the multiuser state):

is:2:initdefault: 
s2:23:wait:/etc/rc2 >/dev/console 2>&1 </dev/console 
co:23:respawn:/etc/gl/conslog 
t1:23:respawn:/etc/getty -s console ttyd1 co_9600 #altconsole
t2:23:off:/etc/getty ttyd2 co_9600 # port 2 
t3:23:off:/etc/getty ttyd3 co_9600 # port 3 
t4:23:off:/etc/getty ttyd4 co_9600 # port 4

The other entries shown above specify the actions necessary to prepare the system to change to the multiuser run level. First, /etc/rc2 is executed. It executes all files in /etc/rc2.d that begin with the letter S, accomplishing (among other things) the following:

init then starts a getty for the console and starts getty on the lines connected to the ports indicated.

At this point, the full multiuser environment is established, and your system is available for users to log in.

Changing Run Levels

To change run levels, the system administrator enters a command that directs init to execute entries in /etc/inittab for a new run level, such as multi, single, or reboot. Then key procedures, such as shutdown, /etc/rc0, and /etc/rc2, are run to initialize the new state.

The new state is reached. If it is state 1 (single-user mode), the system administrator can continue.

Run-level Directories

Run levels 0, 2, and 3 each have a directory of files that are executed in transitions to and from that level. These directories are rc0.d, rc2.d, and rc3.d, respectively. All files in these directories are linked to files in /etc/init.d. The run-level file names look like this:

S00name

or

K00name

The filenames can be split into three parts:

S or K

The first letter defines whether the process should be started (S) or killed (K) upon entering the new run level.

00

The next two characters are a number from 00 to 99. They indicate the order in which the files will be started (S00, S01, S02, etc.) or stopped (K00, K01, K02, etc).

name

The rest of the filename is the /etc/init.d filename to which this file is linked.

For example, the init.d file cron is linked to the rc2.d file and rc0.d file S75cron. When you enter init 2, this file is executed with the start option: sh S75cron start. When you enter init 0, this file is executed with the stop option: sh K70cron stop. This particular shell script executes /sbin/cron when run with the start option and kills the cron process when run with the stop option.

Because these files are shell scripts, you can read them to see what they do. You can modify these files, though it is preferable to add your own since the delivered scripts may change in future releases. To create your own scripts, follow these rules:

Note that it may prove easier to simply copy an existing script from the directory and make appropriate changes.

Going to Single-user Mode From Multiuser Mode

Sometimes you must perform administrative functions, such as backing up the root file system, in single-user mode.

Use the shutdown command. This procedure executes all the files in /etc/rc0.d by calling the /etc/rc0 procedure. The shutdown command does the following things, among others:

The entries for single-user processing in the sample /etc/inittab are:

s1:1:wait:/etc/shutdown -y -iS -g0 >/dev/console 2>&1 </dev/console

There are three recommended ways to start the shutdown to single-user mode:

    You can enter the shutdown -i1 command (recommended). The option -g specifies a grace period between the first warning message and the final message.

    You can enter the single command, which runs a shell script that switches to single-user mode and turns the getty processes off.

    You can enter the init 1 command, which forces the init process to scan the table. The first entry it finds is the s1 entry, and init starts the shutdown processing according to that entry.

Now the system is in the single-user environment, and you can perform the appropriate administrative tasks.

/etc/inittab and Power Off

The following entries in /etc/inittab power off the system: 

s0:06s:wait:/etc/rc0 >/dev/console 2>&1 </dev/console 
of:0:wait:/etc/uadmin 2 0 >/dev/console 2>&1 </dev/console

Always attempt to shut the system down gracefully. You can either enter the powerdown command, the init 0 command, or directly invoke the /etc/shutdown -i0 command.

In either case, the /etc/shutdown and /etc/rc0 procedures are called to clean up and stop all user processes, daemons, and other services and to unmount the file systems. Finally, the /sbin/uadmin command is called, which indicates that the last step (physically removing power from the system) is under firmware control.

Encapsulated PostScript File v.3.0 vs. PostScript File Format

In the course of maintaining your system, you are likely to receive files in various versions of the PostScript format. Following are some of the main differences between the Encapsulated PostScript® File (EPSF) version 3.0 format and PostScript file format:

The EPSF format was designed for importing the PostScript description of a single page or part of a page, such as a figure, into a document, without affecting the rest of the description of that document. EPS code should be encapsulated, i.e., it should not interfere with the PS code that may surround it, and it should not depend on that code.

The EPSF format is usually used for the output from a program that was designed for the preparation of illustrations or figures, (such as the Adobe Illustrator) and as input into desktop publishing programs (such as the Frame Technology's FrameMaker). Most desktop publishing programs can produce the description of a document in the PostScript format that may include the imported EPS code.

For more information about these formats, see the book PostScript Language Reference Manual, Second Edition, Addison-Wesley Publishing Company, Inc., Reading, MA, 1990. You can get several documents on the EPS and PS formats from the Adobe Systems PostScript file server by entering the following at a UNIX prompt, and then following the instructions you get from the file server:

mailps-file-server@adobe.com
Subject:help
Ctrl-D

You can get a description of the EPSF format in the PS format by sending the following message to that file server:

sendDocumentsEPSF2.0.ps

[Previous Section] [Back to Table of Contents] [Next Section]

Send feedback to Technical Publications.

Copyright © 1997, Silicon Graphics, Inc. All Rights Reserved. Trademark Information