Paging and Swapping Devices

Linux stores information about devices used for system paging and swapping activities in a file called /proc/swaps. To display information regarding swap devices, the file can be displayed using either the cat command or the swapon command with the -s option. As shown below, the results are the same using either method.

# cat /proc/swaps

Filename                                Type            Size    Used    Priority
/dev/hdf2                               partition       524152  0       -1
# swapon -s
Filename                                Type            Size    Used    Priority
/dev/hdf2                               partition       524152  0       -1

Here we see the device name, type, size and amount used for each swap device.  In this case we only have the one swap device.

Kernel Parameters

Linux is a parameter driven system. Kernel parameters used for system configuration are found in /proc/sys/kernel, where there is an individual file for each configuration parameter. Since these parameters have a direct effect on system performance, root access is required in order to modify them.

You can view the current configuration parameters by using the cat command on the appropriate file.

# cat msgmax
8192

Occasionally, a prerequisite to a software installation requires the modification of kernel parameters.  Since each parameter file contains a single line of data consisting of either a text string or numeric values, it is often easy to modify a parameter by simply using the echo command:

# echo 2048 > /proc/sys/kernel/msgmax

The aforementioned command will set the value of the msgmax parameter to 2048.  This change will only remain in effect until the system is rebooted.

Linux also provides the sysctl command to modify kernel parameters at runtime. sysctl uses parameter information stored in a file called /etc/sysctl.conf. If, for example, we needed to change the value of the msgmax parameter as was accomplished above, we could accomplish the same thing with the sysctl command like this:

# sysctl -w kernel.msgmax=2048

The next section will introduce some useful utilities that are provided with most Linux releases.  These utilities allow some interactive monitoring of the system.

Server monitoring commands

Some popular commands/utilities for monitoring system resources and tasks managed by the Linux kernel are as follows:

* top: Provides a dynamic real-time view of a running system, including information about system resource usage and a constantly updated list of the processes which are consuming the most resources.  Because it is so useful for administration we will talk quite a bit about top in chapter 8.

* mpstat: Reports activities for each available processor, processor zero being the first one reported.  Global average activities across all processors are also reported.

* iostat: Used for monitoring the load on system input/output devices by observing the time the devices are active compared to the average transfer rate of the device.

* vmstat: Displays information about processes, memory, paging, block IO, and different levels of CPU activity

Interactive Statistics using the top Utility

The top program provides a dynamic real-time view of a running Linux system.  This can be one of the most useful ways to monitor the system as it shows several key statistics on the same page with information on the busiest processes.  It is also one of the few commands which will constantly update.  For more information on the top utility refer to chapter 8.

Displaying Multi-Processor Statistics

The mpstat command displays statistics for each and every processor within the system. The mpstat command accepts interval (delay) and count (repeat) values as parameters.  mpstat will repeat the output count times every interval seconds.  This is very useful for monitoring changes over time.

The following example uses an interval of three seconds and a count of five iterations of the report detail lines. An average summary line is produced at the end of the report.

$ mpstat 3 5

Linux 2.6.5-1.358 (Dell-Linux)  10/18/2004
10:33:26 PM  CPU   %user   %nice %system %iowait    %irq   %soft   %idle    intr/s
10:33:29 PM  all    1.00    0.00    0.33    0.00    0.00    0.00   98.67   1001.99
10:33:32 PM  all    0.33    0.00    0.33    0.00    0.00    0.00   99.33   1007.02
10:33:35 PM  all    0.67    0.00    0.33    0.00    0.00    0.00   99.00   1002.67
10:33:38 PM  all    0.66    0.00    0.33    0.00    0.00    0.00   99.00   1000.33
10:33:41 PM  all    0.67    0.00    0.33    0.00    0.00    0.00   99.00   1005.67
Average:     all    0.67    0.00    0.33    0.00    0.00    0.00   99.00   1003.53

The columns in the report generated by the mpstat command are defined as follows:

* CPU: Either the processor number or the keyword all, which indicates that statistics are calculated as averages among all processors or that there is only one processor in the server

* %user: The percentage of CPU used by user applications

* %nice: The percentage of CPU utilization at the user level with nice priority

* %system: The percentage of CPU used by the system. This does not include the time spent servicing interrupts or softirqs. A softirq is a software interrupt, one of up to 32 software interrupts which can run on multiple CPUs simultaneously.

* %iowait: The percentage of time the system had a pending disk I/O request

* %irq: The percentage of time spent by the CPUs servicing interrupts

* %soft: The percentage of time the processors spent servicing softirqs.

* %idle: The percentage of time that the processors were idle and the system did not have a pending disk I/O request.

* intr/s: The total number of interrupts per second received by the processor(s)