Host and Database Performance Monitoring - Oracle Exadata Recipes: A Problem-Solution Approach - page 406

Download and Install collectl and collectl-utils
The collectl and collectl-utils software is located at collectl.sourceforge.net and
collectl-utils.sourceforge.net , respectively. Download the software from these locations, transfer the files to
each node in your Exadata Database Machine to a directory of your choice, uncompress and untar the contents, and
install them using the INSTALL program as root . Refer to the README files for each utility for more information.
Installing collectl and its utilities will place Perl scripts in /usr/bin on each node. The /usr/bin/collectl
script is the main collectl program and, in this recipe, we will also demonstrate using /usr/bin/colmux to collate
collectl information from multiple Exadata nodes.
Not ■ gnuplot is not installed on the exadata Compute nodes or storage cells, so the installation of collectl-utils
will generate a warning message. You can safely ignore this message, but bear in mind that plotting functionality with
colplot will not be available. to graph collectl output, you can use excel, plot using R scripts, or use the graphing tool
of your choice.
Execute collectl with No Command Line Options
To get a feel for collectl 's output, simply execute collectl while logged in as any user on the nodes that you have
installed the software on. When you execute the collectl program without arguments, the default output will display
CPU utilization statistics, disk I/O statistics, and Ethernet network performance statistics:
[root@cm01dbm01 ~]# collectl
waiting for 1 second sample...
#<−−--CPU[HYPER]-----><−−--------Disks-----------><−−--------Network---------->
#cpu sys inter ctxsw KBRead Reads KBWrit Writes KBIn PktIn KBOut PktOut
0 0 11554 18592 0 0 0 0 1 11 6 8
1 0 12013 19333 0 0 0 0 4 18 2 19
0 0 10767 18487 0 0 824 32 1 8 7 8
1 0 12195 20860 0 0 0 0 0 3 0 3
... Output omitted for brevity
[root@cm01dbm01 ~]#
Running collectl on a Single Node
collectl includes a number of command line arguments and switches to tailor the output to your needs. By default,
collect uses the -scdn command line switch, where s represents the switch arguments, c denotes CPU utilization,
d indicates disk I/O statistics, and n represents Ethernet network information. As you can see, collectl allows you to
report multiple operating system performance statistics on a single line in real time.
Use collectl -showoptions and collectl -x to display the collectl help and command line switches that are
available. Additionally, you can find this information in collectl 's man page.
On Exadata Database Machines, several performance metrics may be important to monitor. With collectl , these
are referred to as subsystems. Some of the more commonly monitored subsystems with c ollectl are CPU utilization,