System Admin Best Practices for Monitoring Linux Systems


System Admin Best Practices for Monitoring Linux Systems

Recipe ID: hsts-r60


Recipe Overview

One of the most important responsibilities a system administrator has, is monitoring their systems. As a system administrator you'll need the ability to find out what is happening on your system at any given time. Whether it's the percentage of system's resources currently used, what commands are being run, or who is logged on. This article will cover how to monitor your system, and in some cases, how to resolve problems that may arise.
When a performance issue arises, there are 4 main areas to consider: CPU, Memory, Disk I/O, and Network. The ability to determine where the bottleneck is can save you a lot of time.

1. System Resources
Being able to monitor the performance of your system is essential. If system resources become to low it can cause a lot of problems. System resources can be taken up by individual users, or by services your system may host such as email or web pages. The ability to know what is happening can help determine whether system upgrades are needed, or if some services need to be moved to another machine.

 

1.1. The top command.
The most common of these commands is top. The top will display a continually updating report of system resource usage.


# top
12:10:49  up 1 day,  3:47,  7 users,  load average: 0.23, 0.19, 0.10
125 processes: 105 sleeping, 2 running, 18 zombie, 0 stopped
CPU states:   5.1% user   1.1% system   0.0% nice   0.0% iowait  93.6% idle
Mem:   512716k av,  506176k used,    6540k free,       0k shrd,   21888k buff
Swap: 1044216k av,  161672k used,  882544k free                  199388k cached

  PID USER     PRI  NI  SIZE  RSS SHARE STAT %CPU %MEM   TIME CPU COMMAND
2330 root      15   0  161M  70M  2132 S     4.9 14.0  1000m   0 X
2605 weeksa    15   0  8240 6340  3804 S     0.3  1.2   1:12   0 kdeinit
3413 weeksa    15   0  6668 5324  3216 R     0.3  1.0   0:20   0 kdeinit
18734 root      15   0  1192 1192   868 R     0.3  0.2   0:00   0 top
1619 root      15   0   776  608   504 S     0.1  0.1   0:53   0 dhclient
1 root      15   0   480  448   424 S     0.0  0.0   0:03   0 init
2 root      15   0     0    0     0 SW    0.0  0.0   0:00   0 keventd
3 root      15   0     0    0     0 SW    0.0  0.0   0:00   0 kapmd
4 root      35  19     0    0     0 SWN   0.0  0.0   0:00   0 ksoftirqd_CPU0
9 root      25   0     0    0     0 SW    0.0  0.0   0:00   0 bdflush
5 root      15   0     0    0     0 SW    0.0  0.0   0:00   0 kswapd
10 root      15   0     0    0     0 SW    0.0  0.0   0:00   0 kupdated
11 root      25   0     0    0     0 SW    0.0  0.0   0:00   0 mdrecoveryd
15 root      15   0     0    0     0 SW    0.0  0.0   0:01   0 kjournald
81 root      25   0     0    0     0 SW    0.0  0.0   0:00   0 khubd
1188 root      15   0     0    0     0 SW    0.0  0.0   0:00   0 kjournald
1675 root      15   0   604  572   520 S     0.0  0.1   0:00   0 syslogd
1679 root      15   0   428  376   372 S     0.0  0.0   0:00   0 klogd
1707 rpc       15   0   516  440   436 S     0.0  0.0   0:00   0 portmap
1776 root      25   0   476  428   424 S     0.0  0.0   0:00   0 apmd
1813 root      25   0   752  528   524 S     0.0  0.1   0:00   0 sshd
1828 root      25   0   704  548   544 S     0.0  0.1   0:00   0 xinetd
1847 ntp       15   0  2396 2396  2160 S     0.0  0.4   0:00   0 ntpd
1930 root      24   0    76    4     0 S     0.0  0.0   0:00   0 rpc.rquotad

The top portion of the report lists information such as the system time, uptime, CPU usage, physical ans swap memory usage, and number of processes. Below that is a list of the processes sorted by CPU utilization.
You can modify the output of top while is is running. If you hit an i, top will no longer display idle processes. Hit i again to see them again. Hitting M will sort by memory usage, S will sort by how long they processes have been running, and P will sort by CPU usage again.
In addition to viewing options, you can also modify processes from within the top command. You can use u to view processes owned by a specific user, k to kill processes, and r to reboot them.
For more in-depth information about processes you can look in the /proc filesystem. In the /proc filesystem you will find a series of sub-directories with numeric names. These directories are associated with the processes ids of currently running processes. In each directory you will find a series of files containing information about the process.
YOU MUST TAKE EXTREME CAUTION TO NOT MODIFY THESE FILES, DOING SO MAY CAUSE SYSTEM PROBLEMS!

 

1.2. The iostat command.
The iostat will display the current CPU load average and disk I/O information. This is a great command to monitor your disk I/O usage.


# iostat
Linux 2.4.20-24.9 (myhost)       12/23/2003

avg-cpu:  %user   %nice    %sys   %idle
62.09    0.32    2.97   34.62

Device:            tps   Blk_read/s   Blk_wrtn/s   Blk_read   Blk_wrtn
dev3-0            2.22        15.20        416    1546846    4799520

For 2.4 kernels the devices is names using the device's major and minor number. In this case the device listed is /dev/hda. To have iostat print this out for you, use the -x.


# iostat -x
Linux 2.4.20-24.9 (myhost)       12/23/2003

avg-cpu:  %user   %nice    %sys   %idle
62.01    0.32    2.97   34.71

Device:  rrqm/s wrqm/s r/s  w/s rsec/s wsec/s rkB/s wkB/s avgrq-sz avgqu-sz await svctm %util
/dev/hdc   0.00   0.00 .00 0.00   0.00   0.00  0.00  0.00     0.00     2.35  0.00  0.00 14.71
/dev/hda   1.13   4.50 .81 1.39  15.18  414  59 23.57    28.24     1.99 63.76 70.48 15.56
/dev/hda1  1.08   3.98 .73 1.27  14.49  42.05  25 21.02    28.22     0.44 21.82  4.97  1.00
/dev/hda2  0.00   0.51 .07 0.12   0.55   5.07  0.27  2.54    30.35     0.97 52.67 61.73  2.99
/dev/hda3  0.05   0.01 .02 0.00   0.14   0.02  0.07  0.01     8.51     0.00 12.55  2.95  0.01

The iostat man page contains a detailed explanation of what each of these columns mean.

 

1.3. The ps command
The ps will provide you a list of processes currently running. There is a wide variety of options that this command gives you.
A common use would be to list all processes currently running. To do this you would use the ps -ef command. (Screen output from this command is too large to include, the following is only a partial output.)


UID        PID  PPID  C STIME TTY          TIME CMD
root         1     0  0 Dec22 ?        00:00:03 init
root         2     1  0 Dec22 ?        00:00:00 [keventd]
root         3     1  0 Dec22 ?        00:00:00 [kapmd]
root         4     1  0 Dec22 ?        00:00:00 [ksoftirqd_CPU0]
root         9     1  0 Dec22 ?        00:00:00 [bdflush]
root         5     1  0 Dec22 ?        00:00:00 [kswapd]
root         6     1  0 Dec22 ?        00:00:00 [kscand/DMA]
root         7     1  0 Dec22 ?        00:01:28 [kscand/Normal]
root         8     1  0 Dec22 ?        00:00:00 [kscand/HighMem]
root        10     1  0 Dec22 ?        00:00:00 [kupdated]
root        11     1  0 Dec22 ?        00:00:00 [mdrecoveryd]
root        15     1  0 Dec22 ?        00:00:01 [kjournald]
root        81     1  0 Dec22 ?        00:00:00 [khubd]
root      1188     1  0 Dec22 ?        00:00:00 [kjournald]
root      1675     1  0 Dec22 ?        00:00:00 syslogd -m 0
root      1679     1  0 Dec22 ?        00:00:00 klogd -x
rpc       1707     1  0 Dec22 ?        00:00:00 portmap
root      1813     1  0 Dec22 ?        00:00:00 /usr/sbin/sshd
ntp       1847     1  0 Dec22 ?        00:00:00 ntpd -U ntp
root      1930     1  0 Dec22 ?        00:00:00 rpc.rquotad
root      1934     1  0 Dec22 ?        00:00:00 [nfsd]
root      1942     1  0 Dec22 ?        00:00:00 [lockd]
root      1943     1  0 Dec22 ?        00:00:00 [rpciod]
root      1949     1  0 Dec22 ?        00:00:00 rpc.mountd
root      1961     1  0 Dec22 ?        00:00:00 /usr/sbin/vsftpd /etc/vsftpd/vsftpd.conf
root      2057     1  0 Dec22 ?        00:00:00 /usr/bin/spamd -d -c -a
root      2066     1  0 Dec22 ?        00:00:00 gpm -t ps/2 -m /dev/psaux
bin       2076     1  0 Dec22 ?        00:00:00 /usr/sbin/cannaserver -syslog -u bin
root      2087     1  0 Dec22 ?        00:00:00 crond
daemon    2195     1  0 Dec22 ?        00:00:00 /usr/sbin/atd
root      2215     1  0 Dec22 ?        00:00:11 /usr/sbin/rcd
weeksa    3414  3413  0 Dec22 pts/1    00:00:00 /bin/bash
weeksa    4342  3413  0 Dec22 pts/2    00:00:00 /bin/bash
weeksa   19121 18668  0 12:58 pts/2    00:00:00 ps -ef

The first column shows who owns the process. The second column is the process ID. The Third column is the parent process ID. This is the process that generated, or started, the process. The forth column is the CPU usage (in percent). The fifth column is the start time, of date if the process has been running long enough. The sixth column is the tty associated with the process, if applicable. The seventh column is the cumulative CPU usage (total amount of CPU time is has used while running). The eighth column is the command itself.
With this information you can see exactly what is running on your system and kill run-away processes, or those that are causing problems.

 

1.4. The vmstat command
The vmstat command will provide a report showing statistics for system processes, memory, swap, I/O, and the CPU. These statistics are generated using data from the last time the command was run to the present. In the case of the command never being run, the data will be from the last reboot until the present.


# vmstat
procs                      memory      swap          io     system      cpu
r  b  w   swpd   free   buff  cache   si   so    bi    bo   in    cs us sy id
0  0  0 181604  17000  26296 201120    0    2     8    24  149     9 61  3 36

The following was taken from the vmstat man page.

FIELD DESCRIPTIONS
Procs
    r: The number of processes waiting for run time.
    b: The number of processes in uninterruptable sleep.
    w: The number of processes swapped out but otherwise runnable.  This
       field is calculated, but Linux never desperation swaps.

Memory
    swpd: the amount of virtual memory used (kB).
    free: the amount of idle memory (kB).
    buff: the amount of memory used as buffers (kB).

Swap
    si: Amount of memory swapped in from disk (kB/s).
    so: Amount of memory swapped to disk (kB/s).

IO
    bi: Blocks sent to a block device (blocks/s).
    bo: Blocks received from a block device (blocks/s).

System
    in: The number of interrupts per second, including the clock.
    cs: The number of context switches per second.

CPU
    These are percentages of total CPU time.
    us: user time
    sy: system time
    id: idle time

 

1.5. The lsof command
The lsof command will print out a list of every file that is in use. Since Linux considers everything a file, this list can be very long. However, this command can be useful in diagnosing problems. An example of this is if you wish to unmount a filesystem, but you are being told that it is in use. You could use this command and grep for the name of the filesystem to see who is using it.
Or suppose you want to see all files in use by a particular process. To do this you would use lsof -p -processid-.

 

1.6. Finding More Utilities
There are lots of online resources for learning what tools are out there and how to do a number of tasks with them.

 

2. Filesystem Usage
Many reports are currently talking about how cheap storage has gotten, but if you're like most of us it isn't cheap enough. Most of us have a limited amount of space, and need to be able to monitor it and control how it's used.

 

2.1. The df command
The df is the simplest tool available to view disk usage. Simply type in df and you'll be shown disk usage for all your mounted filesystems in 1K blocks


user@server:~> df
Filesystem           1K-blocks      Used Available Use% Mounted on
/dev/hda3              5242904    759692   4483212  15% /
tmpfs                   127876         8    127868   1% /dev/shm
/dev/hda1               127351     33047     87729  28% /boot
/dev/hda9             10485816     33508  10452308   1% /home
/dev/hda8              5242904    932468   4310436  18% /srv
/dev/hda7              3145816     32964   3112852   2% /tmp
/dev/hda5              5160416    474336   4423928  10% /usr
/dev/hda6              3145816    412132   2733684  14% /var

You can also use the -h to see the output in "human-readable" format. This will be in K, Megs, or Gigs depending on the size of the filesystem. Alternately, you can also use the -B to specify block size.
In addition to space usage, you could use the -i option to view the number of used and available inodes.


user@server:~> df -i
Filesystem            Inodes   IUsed   IFree IUse% Mounted on
/dev/hda3                  0       0       0    -  /
tmpfs                  31969       5   31964    1% /dev/shm
/dev/hda1              32912      47   32865    1% /boot
/dev/hda9                  0       0       0    -  /home
/dev/hda8                  0       0       0    -  /srv
/dev/hda7                  0       0       0    -  /tmp
/dev/hda5             656640   26651  629989    5% /usr
/dev/hda6                  0       0       0    -  /var

 

2.2. The du command
Now that you know how much space has been used on a filesystem how can you find out where that data is? To view usage by a directory or file you can use du. Unless you specify a filename du will act recursively. For example:


user@server:~> du file.txt
1300    file.txt

Or like the df I can use the -h and get the same output in "human-readable" form.


user@server:~> du -h file.txt
1.3M     file.txt

Unless you specify a filename du will act recursively.


user@server:~> du -h /usr/local
4.0K    /usr/local/games
16K     /usr/local/include/nessus/net
180K    /usr/local/include/nessus
208K    /usr/local/include
62M     /usr/local/lib/nessus/plugins/.desc
97M     /usr/local/lib/nessus/plugins
164K    /usr/local/lib/nessus/plugins_factory
97M     /usr/local/lib/nessus
12K     /usr/local/lib/pkgconfig
2.7M    /usr/local/lib/ladspa
104M    /usr/local/lib
112K    /usr/local/man/man1
4.0K    /usr/local/man/man2
4.0K    /usr/local/man/man3
4.0K    /usr/local/man/man4
16K     /usr/local/man/man5
4.0K    /usr/local/man/man

If you just want a summary of that directory you can use the -s option.


user@server:~> du -hs /usr/local
210M    /usr/local

 

2.3. Quotas
You can find more information about quotas online.

 

3. Monitoring Users
Just because you're paranoid doesn't mean they AREN'T out to get you... Source Unknown
From time to time there are going to be occasions where you will want to know exactly what people are doing on your system. Maybe you notice that a lot of RAM is being used, or a lot of CPU activity. You are going to want to see who is on the system, what they are running, and what kind of resources they are using.

 

3.1. The who command
The easiest way to see who is on the system is to do a who or w. The --> who is a simple tool that lists out who is logged --> on the system and what port or terminal they are logged on at.


user@server:~>  who
bjones   pts/0        May 23 09:33
wally    pts/3        May 20 11:35
aweeks   pts/1        May 22 11:03
aweeks   pts/2        May 23 15:04

 

3.2. The ps command -again!
In the previous section we can see that user aweeks is logged onto both pts/1 and pts/2, but what if we want to see what they are doing? We could to a ps -u aweeks and get the following output


user@server:~> ps -u aweeks

20876 pts/1    00:00:00 bash
20904 pts/2    00:00:00 bash
20951 pts/2    00:00:00 ssh
21012 pts/1    00:00:00 ps

From this we can see that the user is doing a ps ssh.
This is a much more consolidated use of the ps than discussed previously.

 

3.3. The w command
Even easier than using the who and ps -u commands is to use the w. w will print out not only who is on the system, but also the commands they are running.


user@server:~> w
aweeks   :0        09:32   ?xdm?  30:09   0.02s -:0
aweeks   pts/0     09:33    5:49m  0.00s  0.82s kdeinit: kded
aweeks   pts/2     09:35    8.00s  0.55s  0.36s vi sag-0.9.sgml
aweeks   pts/1     15:03   59.00s  0.03s  0.03s /bin/bash

From this we can see that I have a kde session running, I'm working in this document :-), and have another terminal open sitting idle at a bash prompt.

Additional Linux Resources

Here is a list of resources for learning Linux:
Resources for System Administrators

Resources for Linux Kernel Programmers

Linux File System Dictionary
Comprehensive Review of How Linux File and Directory System Works

Hands-on Linux classes

Linux Operating System Distributions

 


Private and Custom Tutoring

We provide private tutoring classes online and offline (at our DC site or your preferred location) with custom curriculum for almost all of our classes for $50 per hour online or $75 per hour in DC. Give us a call or submit our private tutoring registration form to discuss your needs.


View Other Classes!