20 KiB
** Linchenguang is translating this essay
Collectl: An Advanced All-in-One Performance Monitoring Tool for Linux
The most important duty of a Linux system administrator is to make sure the system he/she is administering is in a very good condition. There are many tools available for a Linux system admins that can help to monitor and display processes in a system such as top and htop, but none of these tools can compete with collectl.
collectl is a very nice feature rich command-line utility that can be used to collect performance data that describes the current system status. Unlike most of the other monitoring tools, collectl does not focus in a limited number of system metrics, instead it can gather information on many different types of system resources such as cpu, disk, memory, network, sockets, tcp, inodes, infiniband, lustre, memory, nfs, processes, quadrics, slabs and buddyinfo.
A very good thing about using collectl is that it can also play the role of utilities that are designed with only a specific purpose such as top, ps, iotop and many others. What are some features that make colleclt a useful tool?
After doing a lot of research I have compiled a list with some of the most important features of the collectl command-line utility for you guys.
Collectl Features
- It can run interactively, as a daemon or both.
- It can display the output in many formats.
- It has the ability to monitor almost any subsystem.
- It can play the role of many other utilities such as ps, top, iotop, vmstat.
- It has the ability to record and playback the captured data.
- It can export the data in various file formats. (this is very useful when you want to analyse the data with external tools).
- It can run as a service to monitor remote machines or an entire server cluster.
- It can display the data in the terminal, write to a file or a socket.
How to Install collectl in Linux
The collectl utility runs on all Linux distributions, the only thing it requires to run is perl, so make sure that you have Perl installed in your machine before installing collectl in your machine.
On Debian/Ubuntu/Linux Mint
The following command can be used to install the collectl utility in Debian based machines such as Ubuntu.
On Debian/Ubuntu/Linux Mint
The following command can be used to install the collectl utility in Debian based machines such as Ubuntu.
$ sudo apt-get install collectl
On RHEL/CentOS/Fedora
If you are using Red Hat based distro, you can easily grab it from the repos with the yum command.
# yum install collectl
Some Practical Examples Of The collectl Utility
Once the installation of the collectl tool is finished, you can easily run it from the terminal, even without any option. The following command will display information on cpu, disk and network stats in a very short and human readable format.
# collectl
waiting for 1 second sample...
#
#cpu sys inter ctxsw KBRead Reads KBWrit Writes KBIn PktIn KBOut PktOut
13 5 790 1322 0 0 92 7 4 13 0 5
10 2 719 1186 0 0 0 0 3 9 0 4
12 0 753 1188 0 0 52 3 2 5 0 6
13 2 733 1063 0 0 0 0 1 1 0 1
25 2 834 1375 0 0 0 0 1 1 0 1
28 2 870 1424 0 0 36 7 1 1 0 1
19 3 949 2271 0 0 44 3 1 1 0 1
17 2 809 1384 0 0 0 0 1 6 0 6
16 2 732 1348 0 0 0 0 1 1 0 1
22 4 993 1615 0 0 56 3 1 2 0 3
As you guys can see from the above output displayed in the terminal screen, it is very easy to work with the system metrics values present in the command output because it appears on a single line.
When the collectl utility is executed without any option it displays information about the following subsystems:
- cpu
- disks
- network
Note: In our case, a subsystem is every type of a system resource that can be measured.
You can also display statistics for all subsystems except slabs by combining the command with the –all option like shown below.
# collectl --all
waiting for 1 second sample...
#
#cpu sys inter ctxsw Cpu0 Cpu1 Free Buff Cach Inac Slab Map Fragments KBRead Reads KBWrit Writes KBIn PktIn KBOut PktOut IP Tcp Udp Icmp Tcp Udp Raw Frag Handle Inodes Reads Writes Meta Comm
16 3 817 1542 430 390 1G 175M 1G 683M 193M 1G nsslkjjebbk 0 0 24 3 1 1 0 1 0 0 0 0 623 0 0 0 8160 240829 0 0 0 0
11 1 745 1324 316 426 1G 175M 1G 683M 193M 1G nsslkjjebbk 0 0 0 0 0 3 0 2 0 0 0 0 622 0 0 0 8160 240828 0 0 0 0
15 2 793 1683 371 424 1G 175M 1G 683M 193M 1G ssslkjjebbk 0 0 0 0 1 1 0 1 0 0 0 0 622 0 0 0 8160 240829 0 0 0 0
16 2 872 1875 427 446 1G 175M 1G 683M 193M 1G ssslkjjebbk 0 0 24 3 1 1 0 1 0 0 0 0 622 0 0 0 8160 240828 0 0 0 0
24 2 842 1383 473 368 1G 175M 1G 683M 193M 1G ssslkjjebbk 0 0 168 6 1 1 0 1 0 0 0 0 622 0 0 0 8160 240828 0 0 0 0
27 3 844 1099 478 365 1G 175M 1G 683M 193M 1G nsslkjjebbk 0 0 0 0 1 6 1 9 0 0 0 0 622 0 0 0 8160 240828 0 0 0 0
26 5 823 1238 396 428 1G 175M 1G 683M 193M 1G ssslkjjebbk 0 0 0 0 2 11 3 9 0 0 0 0 622 0 0 0 8160 240828 0 0 0 0
15 1 753 1276 361 391 1G 175M 1G 683M 193M 1G ssslkjjebbk 0 0 40 3 1 2 0 3 0 0 0 0 623 0 0 0 8160 240829 0 0 0 0
But, how do you monitor the cpu usage with the help of the utility? The ‘-s‘ option should be used to controls which subsystem data is to be collected or played back.
For example the following command can be used to monitor the summary of the cpu usage.
# collectl -sc
waiting for 1 second sample...
#
#cpu sys inter ctxsw
15 2 749 1155
16 3 772 1445
14 2 793 1247
27 4 887 1292
24 1 796 1258
16 1 743 1113
15 1 743 1179
14 1 706 1078
15 1 764 1268
What happens when you combine the command with “scdn“? The best way to learn how to use command-line tools is to practice as much as possible, so run the following command in your terminal and see what is going to happen.
# collectl -scdn
waiting for 1 second sample...
#
#cpu sys inter ctxsw KBRead Reads KBWrit Writes KBIn PktIn KBOut PktOut
25 4 943 3333 0 0 0 0 1 1 0 2
27 3 825 2910 0 0 0 0 1 1 0 1
27 5 886 2531 0 0 0 0 0 0 0 1
20 4 872 2406 0 0 0 0 1 1 0 1
26 1 854 2091 0 0 20 2 1 1 0 1
39 4 1004 3398 0 0 0 0 2 8 3 6
41 6 955 2464 0 0 40 3 1 2 0 3
25 7 890 1609 0 0 0 0 1 1 0 1
16 2 814 1165 0 0 796 43 2 2 0 2
14 1 779 1383 0 0 48 6 1 1 0 1
11 2 795 1285 0 0 0 0 2 14 1 14
You can easily understand that the default option is “cdn“, it stands for cpu, disks and network data. The result of the command is the same with the output of “collectl -scn“
If you want to collect data about the memory, use the following command.
# collectl -sm
waiting for 1 second sample...
#
#Free Buff Cach Inac Slab Map
1G 177M 1G 684M 193M 1G
1G 177M 1G 684M 193M 1G
1G 177M 1G 684M 193M 1G
1G 177M 1G 684M 193M 1G
1G 177M 1G 684M 193M 1G
1G 177M 1G 684M 193M 1G
1G 177M 1G 684M 193M 1G
1G 177M 1G 684M 193M 1G
The above output is very useful when you want to get some detailed information on your memory usage, free memory and other important stuff for the performance of your system.
How about some data on tcp? Use the following command to do it.
# collectl -st
waiting for 1 second sample...
#
# IP Tcp Udp Icmp
0 0 0 0
0 0 0 0
0 0 0 0
0 0 0 0
0 0 0 0
0 0 0 0
0 0 0 0
0 0 0 0
0 0 0 0
0 0 0 0
0 0 0 0
After you have gained some experience you can easily combine options to get the results you want. For example you can combine the “t” for tcp and “c” for cpu. The following command does that.
# collectl -stc
waiting for 1 second sample...
#
#cpu sys inter ctxsw IP Tcp Udp Icmp
23 8 961 3136 0 0 0 0
24 5 916 3662 0 0 0 0
21 8 848 2408 0 0 0 0
30 10 916 2674 0 0 0 0
38 3 826 1752 0 0 0 0
31 3 820 1408 0 0 0 0
15 5 781 1335 0 0 0 0
17 3 802 1314 0 0 0 0
17 3 755 1218 0 0 0 0
14 2 788 1321 0 0 0 0
It is hard for us humans to remember all the available options so I am posting the summary list of subsystems supported by the tool.
- b – buddy info (memory fragmentation)
- c – CPU
- d – Disk
- f – NFS V3 Data
- i – Inode and File System
- j – Interrupts
- l – Lustre
- m – Memory
- n – Networks
- s – Sockets
- t – TCP
- x – Interconnect
- y – Slabs (system object caches)
A very important piece of data for a system administrator or a Linux user is the data collected on disk usage. The following command will help you to monitor the disk usage.
# collectl -sd
waiting for 1 second sample...
#
#KBRead Reads KBWrit Writes
0 0 0 0
0 0 0 0
0 0 92 7
0 0 0 0
0 0 36 3
0 0 0 0
0 0 0 0
0 0 100 7
0 0 0 0
You can also use the “-sD” option to collect data on individual disks, but you have to know that information on total disks will not be reported.
# collectl -sD
waiting for 1 second sample...
# DISK STATISTICS (/sec)
# Pct
#Name KBytes Merged IOs Size KBytes Merged IOs Size RWSize QLen Wait SvcTim Util
sda 0 0 0 0 52 11 2 26 26 1 8 8 1
sda 0 0 0 0 0 0 0 0 0 0 0 0 0
sda 0 0 0 0 24 0 2 12 12 0 0 0 0
sda 0 0 0 0 152 0 4 38 38 0 0 0 0
sda 0 0 0 0 192 45 3 64 64 1 20 20 5
sda 0 0 0 0 204 0 2 102 102 0 0 0 0
sda 0 0 0 0 0 0 0 0 0 0 0 0 0
sda 0 0 0 0 116 26 3 39 38 1 16 16 4
sda 0 0 0 0 0 0 0 0 0 0 0 0 0
sda 0 0 0 0 0 0 0 0 0 0 0 0 0
sda 0 0 0 0 32 5 3 11 10 1 16 16 4
sda 0 0 0 0 0 0 0 0 0 0 0 0 0
You can also use other detail subsystems to collect detailed data. The following is a list of the detail subsystems.
- C – CPU
- D – Disk
- E – Environmental data (fan, power, temp), via ipmitool
- F – NFS Data
- J – Interrupts
- L – Lustre OST detail OR client Filesystem detail
- N – Networks
- T – 65 TCP counters only available in plot format
- X – Interconnect
- Y – Slabs (system object caches)
- Z – Processes
There are many available options in the collectl utility, but there is not enough time and space to cover them all in only one article. However it is worth mentioning and teaching how to use the utility as top and ps.
It is very easy to make collectl work as the top utility, just run the following command in your terminal and you will see the similar output the top tool gives you when it is executed in your Linux system.
# collectl --top
# TOP PROCESSES sorted by time (counters are /sec) 13:11:02
# PID User PR PPID THRD S VSZ RSS CP SysT UsrT Pct AccuTime RKB WKB MajF MinF Command
^COuch!tecmint 20 1 40 R 1G 626M 0 0.01 0.14 15 28:48.24 0 0 0 109 /usr/lib/firefox/firefox
3403 tecmint 20 1 40 R 1G 626M 1 0.00 0.20 20 28:48.44 0 0 0 600 /usr/lib/firefox/firefox
5851 tecmint 20 4666 0 R 17M 13M 0 0.02 0.06 8 00:01.28 0 0 0 0 /usr/bin/perl
1682 root 20 1666 2 R 211M 55M 1 0.02 0.01 3 03:10.24 0 0 0 95 /usr/bin/X
3454 tecmint 20 3403 8 S 216M 45M 1 0.01 0.02 3 01:23.32 0 0 0 0 /usr/lib/firefox/plugin-container
4658 tecmint 20 4657 3 S 207M 17M 1 0.00 0.02 2 00:08.23 0 0 0 142 gnome-terminal
2890 tecmint 20 2571 3 S 340M 68M 0 0.00 0.01 1 01:19.95 0 0 0 0 compiz
3521 tecmint 20 1 24 S 710M 148M 1 0.01 0.00 1 01:47.84 0 0 0 0 skype
1 root 20 0 0 S 3M 2M 0 0.00 0.00 0 00:02.57 0 0 0 0 /sbin/init
2 root 20 0 0 S 0 0 1 0.00 0.00 0 00:00.00 0 0 0 0 kthreadd
3 root 20 2 0 S 0 0 0 0.00 0.00 0 00:00.60 0 0 0 0 ksoftirqd/0
5 root 0 2 0 S 0 0 0 0.00 0.00 0 00:00.00 0 0 0 0 kworker/0:0H
7 root 0 2 0 S 0 0 0 0.00 0.00 0 00:00.00 0 0 0 0 kworker/u:0H
8 root RT 2 0 S 0 0 0 0.00 0.00 0 00:04.42 0 0 0 0 migration/0
9 root 20 2 0 S 0 0 0 0.00 0.00 0 00:00.00 0 0 0 0 rcu_bh
10 root 20 2 0 R 0 0 0 0.00 0.00 0 00:02.22 0 0 0 0 rcu_sched
11 root RT 2 0 S 0 0 0 0.00 0.00 0 00:00.05 0 0 0 0 watchdog/0
12 root RT 2 0 S 0 0 1 0.00 0.00 0 00:00.07 0 0 0 0 watchdog/1
13 root 20 2 0 S 0 0 1 0.00 0.00 0 00:00.73 0 0 0 0 ksoftirqd/1
14 root RT 2 0 S 0 0 1 0.00 0.00 0 00:01.96 0 0 0 0 migration/1
16 root 0 2 0 S 0 0 1 0.00 0.00 0 00:00.00 0 0 0 0 kworker/1:0H
17 root 0 2 0 S 0 0 1 0.00 0.00 0 00:00.00 0 0 0 0 cpuset
And now last but not least, to use the collectl utility as the ps tool run the following command in your terminal. You will get information about processes in your system the same way as you do when you run the “ps” command in your terminal.
# collectl -c1 -sZ -i:1
waiting for 1 second sample...
### RECORD 1 >>> tecmint-vgn-z13gn <<< (1397979716.001) (Sun Apr 20 13:11:56 2014) ###
# PROCESS SUMMARY (counters are /sec)
# PID User PR PPID THRD S VSZ RSS CP SysT UsrT Pct AccuTime RKB WKB MajF MinF Command
1 root 20 0 0 S 3M 2M 0 0.00 0.00 0 00:02.57 0 0 0 0 /sbin/init
2 root 20 0 0 S 0 0 1 0.00 0.00 0 00:00.00 0 0 0 0 kthreadd
3 root 20 2 0 S 0 0 0 0.00 0.00 0 00:00.60 0 0 0 0 ksoftirqd/0
5 root 0 2 0 S 0 0 0 0.00 0.00 0 00:00.00 0 0 0 0 kworker/0:0H
7 root 0 2 0 S 0 0 0 0.00 0.00 0 00:00.00 0 0 0 0 kworker/u:0H
8 root RT 2 0 S 0 0 0 0.00 0.00 0 00:04.42 0 0 0 0 migration/0
9 root 20 2 0 S 0 0 0 0.00 0.00 0 00:00.00 0 0 0 0 rcu_bh
10 root 20 2 0 S 0 0 0 0.00 0.00 0 00:02.24 0 0 0 0 rcu_sched
11 root RT 2 0 S 0 0 0 0.00 0.00 0 00:00.05 0 0 0 0 watchdog/0
12 root RT 2 0 S 0 0 1 0.00 0.00 0 00:00.07 0 0 0 0 watchdog/1
13 root 20 2 0 S 0 0 1 0.00 0.00 0 00:00.73 0 0 0 0 ksoftirqd/1
14 root RT 2 0 S 0 0 1 0.00 0.00 0 00:01.96 0 0 0 0 migration/1
16 root 0 2 0 S 0 0 1 0.00 0.00 0 00:00.00 0 0 0 0 kworker/1:0H
17 root 0 2 0 S 0 0 1 0.00 0.00 0 00:00.00 0 0 0 0 cpuset
18 root 0 2 0 S 0 0 1 0.00 0.00 0 00:00.00 0 0 0 0 khelper
19 root 20 2 0 S 0 0 0 0.00 0.00 0 00:00.00 0 0 0 0 kdevtmpfs
20 root 0 2 0 S 0 0 0 0.00 0.00 0 00:00.00 0 0 0 0 netns
21 root 20 2 0 S 0 0 0 0.00 0.00 0 00:00.00 0 0 0 0 bdi-default
22 root 0 2 0 S 0 0 0 0.00 0.00 0 00:00.00 0 0 0 0 kintegrityd
I am very sure that many Linux system administrators will like this tool and will feel its power when using it to the fullest. If you like to advance your knowledge about collectl to the next level refer to its manual pages and keep practicing.
Just type the following command in your terminal and start reading.
# man collectl
Reference Links
via: http://www.tecmint.com/linux-performance-monitoring-with-collectl-tool/