Softpanorama (slightly skeptical) Open Source Software Educational Society

May the source be with you, but remember the KISS principle ;-)

Process Scheduling

Solaris offers typical UNIX commands for process scheduling. Solaris 10 adds SMF facility for starting stopping processes that gives you capability to take extra advantage of resolving existing dependencies during starting or stopping of a particular process.

As in most Unixes crontab and at commands are used in Solaris for automatic execution of executables and scripts  at a scheduled time. In addition you can use the nice and the renice commands to alter the priority of running processes.

The crontab Command

The crontab files are stored in the /var/spool/cron/crontab directory. Each crontab command represents a scheduled task, which is known as a cron job. Root and any user who has permissions to schedule events has his or her own crontab file.

The structure of crontab did not changes from the times it was introduced. It contains six fields:

1. Minute Minute of the hour (0-59)
2. Hour Hour of the day (0-23)
Day Day of the month (1-31)
4. Month Month of the year (1-12)
5. Weekday Day of the week(0-6; 0=Sunday)
6. Command The command to execute

The cron daemon can be controlled using  /etc/default/cron file. For example to log the activities of the cron command, ensure the presence of the following entry:

CRONLOG=YES

This entry logs the activity of cron in the /var/cron/log file. To disable logging, modify this entry as follows:

CRONLOG=NO

You need to restart the cron to reread the file.

You can control users' access to the crontab command by using the /etc/cron.d/cron.allow and the /etc.cron.d/cron.deny files. The /etc/cron.d/cron.allow file consists of a list of users who are allowed to use the crontab command. This file does not exist by default. A user's access to the crontab command is confirmed after a search for this file. If the file does not exist, the /etc/cron.d/cron.deny file, which contains a list of users whose access to the crontab command is prohibited, is searched.

In Solaris by default the cron.deny file prohibits crontab use by the following system users:

• daemon

• bin

• smtp

• nuucp

• listen

• nobody

• noaccess

If both the files cron.allow and cron.deny do not exist, all users are denied access to crontab except the superuser.

To create a crontab file, use the –e option of the crontab command followed by the file name, for example:

# crontab –e joeuser

The –e option is also used to edit an existing crontab file.

The –l option lists a crontab file. The following example displays the lp crontab file used for printing.

# crontab –l lp 

The –r option is used to delete the crontab file.  

The at command

The at command schedules a single system event only once, unlike the crontab command, which can schedule multiple system events repetitively at specified periods. The at command enables users to create, display, and remove their own at jobs.

To schedule at jobs, type at followed by the time. After you type this, a prompt <at> is displayed. Specify the command or script that has to execute at the specified time. The time is mentioned in minutes, hours, days, and months.

The following example shows how to remove the files from the home directory of user Jency at 9.55 p.m. on August 2.

$ at 9:55PM
at> rm /export/home/jency
at> rm /albert
at> mkdir / albert1
at> <EOT>
commands will be executed using /sbin/sh
job 996807300.a at Thu Aug  2 21:55:00 2001
$

Each at job is given a job ID and has a suffix of .a, which identifies it is an at job. When more than one job is scheduled for execution, it is queued as per the time schedule. To display this queue, use the atq command.

# atq
 Rank     Execution Date     Owner     Job         Queue   Job Name
  1st   Aug  2, 2001 21:55   root    996807300.a     a     stdin
  2nd   Aug 15, 2001 12:00   root    997894800.a     a     stdin
#

Options:

• -m Sends mail to the user after execution of the job.
• -r Removes at jobs.
• -l Displays information about the execution time.

$ at -l
user = root     996807300.a     Thu Aug  2 21:55:00 2001
user = root     997894800.a     Wed Aug 15 12:00:00 2001
$

The at –r command followed by the job ID removes the at job with the specified job ID.

# at –r <job id>

nice and renice commands

You can use the nice command to alter the access priority of processes. The nice number for a process can range from 0 to +40 with 0 indicating the highest priority and 40 indicating the lowest priority. To change the priority order of a process by using the nice command, change the process's nice number. For example, to increase the priority of a process by 10 units, decrease the nice number by 10. The higher the value of nice, the lower is its execution priority.

Both the standard version and the /usr/bin/nice version of the nice command in Solaris use the following syntax:

/usr/bin/nice -[+ | -n] command_name

In the /usr/bin version of the nice command, the value to add or subtract from the nice number is preceded by a hyphen. For example, to decrease the nice number from 30 to 24, use the following syntax. Note that if no number is specified, the nice number is increased by 10 by default, which decreases the priority value by 10 units.

/usr/bin/nice --6 command_name
 

Similarly, to increase the nice number from 24 to 30, use the following syntax:

/usr/bin/nice -6 command_name
 

You can also change the priority of a process while the process is running. Solaris provides the renice command, which you use to change the priority of an executing process. The renice command takes the PID of the process as the operand. The command uses the following syntax:

renice [-n priority_change] PID
 

The –n option in the preceding syntax defines the number of units by which to increase or decrease the priority of the running process. By default, all the processes running on the system are assigned a nice value equal to 20. Note that after altering the priority of the running process, the new nice value is 20 +/- the priority change. This value can range from 0 through 39. PID is the process ID for which the priority has to be changed.

For example, to decrease the priority of the process for PID 324 by 5 during runtime, use the following command:

renice –n 5 –p 324
 

In the preceding command, the new nice value for the process will be 15.

You also can use the renice command to change the priority of processes belonging to a particular user. For example, to increase the execution priority of processes belonging to the user David by 8 units, use the following command:

renice –n –8 –u David

The priocntl command

The Solaris provides the priocntl command, which you use to change the scheduling behavior of a process. The command displays or sets the priority of the processes. You can also use the priocntl command to display the current configuration information of the process scheduler.

Table below lists the options of the priocntl command:

• -l Lists currently loaded scheduling classes.
• -d Displays the scheduling parameters of a process.
• -e Creates a process.
• -p Changes the priority of an existing process.  

The priocntl -l command displays a list of the currently loaded scheduling classes. A sample output of the command follows.

# priocntl -l
CONFIGURED CLASSES
==================
SYS (System Class)
TS (Time Sharing)
        Configured TS User Priority Range: -60 through 60
IA (Interactive)
        Configured IA User Priority Range: -60 through 60
#

# priocntl -d -i pid 1
TIME SHARING PROCESSES:
    PID    TSUPRILIM    TSUPRI
     1            0                0

# priocntl -e -c TS -p 10 find / -name core -print

Truss

Solaris offers the truss utility that you use to track processes running on a system. The truss utility is similar to the trace utility of Solaris 4.x. The truss utility reports the following information about processes:

• System calls made by a process, including arguments and return values

• Signals received by the process

• Machine faults encountered by the process

You use the truss utility to debug problems with processes. Although truss is not a debugging utility, it helps you identify problems a process encounters.

You can use the truss utility to track any executable command or a currently running process by using the PID value of the process. The truss utility tracks the child processes until the process exits. The truss utility uses the following syntax to track processes that are running on the system:

truss –aef –p PID

where PID is the process ID of a currently running process.

To use truss with an executable command, use the following syntax:

truss –aef <command>

Table below lists the commonly used options of the truss command and their descriptions. For a complete list of options, refer to the man pages for the truss command.

By default, the truss utility dumps the output to the stderr file. You can save the output of the truss utility by using the –o option with the truss command. For example, to save the output of the truss command to the trussoutput.out file use the following command:

truss –aef –o /tmp/trussoutput.out –p PID

Summary

The multitasking, multiprocessing, and multithreading design of Solaris enables it to provide enhanced performance for key enterprise applications along with stability and accuracy. A process is a running instance of a program. For example, a browser window running on your workstation is a process of the browser program.

Most processes in the system are created by fork system calls. The fork system call makes a copy (child process) of the calling process (parent process) in a new address space in the virtual memory. The child process continues to execute on the CPU until it completes. On completion, the child process returns the resources to the system. A process during its lifetime can exist in any of the following states: Init, Run, Sleep, and Zombie.

Processes running on a system affect the performance of the system because processes consume system resources, such as CPU time and memory. Therefore, it is important that you manage the processes running on the system. Managing the processes running on the system involves monitoring processes, determining processor usage, changing process priorities, and terminating processes

Last modified: February 28, 2008