Monitor Systems for Vital Characteristics

Install pcp and pcp-system-tools.

Enable and start the pmcd and pmlogger services.

Take a baseline of the kernel.all.load metric for 10 seconds and put this into the file /home/cloud_user/kernel.all.load.txt.

You can do this using the pmval or pmrep command:

pmval -T 10s kernel.all.load > /home/cloud_user/kernel.all.load.txt

Or:

pmrep -T 10s kernel.all.load > /home/cloud_user/kernel.all.load.txt

Take a baseline of the disk.partitions.total_rawactive metric for 10 seconds and put this into the file /home/cloud_user/disk.partitions.total_rawactive.txt.

You can do this using the pmval or pmrep command:

pmval -T 10s disk.partitions.total_rawactive > /home/cloud_user/disk.partitions.total_rawactive.txt

Or:

pmrep -T 10s disk.partitions.total_rawactive > /home/cloud_user/disk.partitions.total_rawactive.txt

By now, pmlogger has been running for a few minutes. Generate some load so that we can look at it in the archive.

Before and after each of the commands that generate load, make a note of the system time. You can do so using the command:

date

Generate some CPU load

Run the following command to generate some CPU load for 1 minute:

date && timeout -sHUP 1m openssl speed

Generate some disk I/O

Run the following command to generate some disk I/O:

date && fallocate -l 1G /home/cloud_user/bigfile && shred -zvu -n 1 /home/cloud_user/bigfile

Make a note of the start and end times from the commands above. We'll need them to know when to look for the increases in resource usages.

Get the pcp archive file:

pcp | grep logger

Look in the archive log directory and make note of the archive files:

ls -lh /var/log/pcp/pmlogger/ip-10-0-1-10.ec2.internal/

Depending on how long you've taken to do these tasks, the archive log may have rolled over to a new file. The format of the filename is YYYYMMDD.HH.MM. Using your notes of when you ran the CPU and disk I/O commands, determine which file to use.

Display the kernel.all.load values from the selected archive log in 1 minute increments:

Note: You can use pmval or pmrep here, with these particular metrics, I find pmrep to be easier to read.

pmrep -t 1m -a /var/log/pcp/pmlogger/ip-10-0-1-10.ec2.internal/<FILE> kernel.all.load

Display the disk.partitions.total_rawactive values from the selected archive log in 1 minute increments:

pmrep -t 1m -a /var/log/pcp/pmlogger/ip-10-0-1-10.ec2.internal/<FILE> disk.partitions.total_rawactive