On the oracle database machine, oswatcher is installed during setup time, both on the database nodes and the exadata cells. This utility collects linux operating system level statistics, which comes in very handy when troubleshooting operating system level issues. The data is collected in text files. There is a Java based utility (OSWG) provided by oracle support to graph the contents of these files, however that utility does not work on the oswatcher files generated on exadata.
Here is a python script that can graph the cpu used from the mpstat information that oswatcher captures. It has been tested on new oswatcher files on an x3-2. You need to first install a python environment that has the “numpy” and “matplotlib” modules installed.
Install a Python Virtualenv.
If you create multiple applications using Python and end up using different versions, it is easier to maintain different virtualenv’s. You can create a python virtualenv as shown below (On ubuntu linux).
curl -O https://pypi.python.org/packages/source/v/virtualenv/virtualenv-1.9.1.tar.gz
tar -xzvf virtualenv-1.9.1.tar.gz
python virtualenv.py ../p273env2
pip install numpy
sudo apt-get install libfreetype6-dev
pip install matplotlib
Now that you have a python environment, with your required libraries, you can go ahead and execute the script as shown below.
The oswatcher files in /opt/oracle/oswatcher are .bz2 files and there will be one file per hour per day. Copy the mpstat .bz2 files into a directory and use bunzip2 to unzip them. In this example let us say that the directory name is /u01/oswatcher/mpstat/tmp
You can now run the script as shown below
python parseoswmp.py /u01/oswatcher/mpstat/tmp
python parseoswmp.py /u01/oswatcher/mpstat/tmp '06/14/2013 05:00:00 AM' '06/14/2013 07:00:00 AM'
The first command will graph the cpu usage for the entire time range in all those files and the second command graphs the cpu information for the date and time range you have specified.
It creates a file in the current directory, named oswmpstat.png, which has the graph.
You can find the full script here.
You can find a sample output graph here.