sar disk io output

loadedmind · 01-03-2017, 11:35 AM

Hello all. As I'm sure most of you have something similar setup, I'm using the sysstat suite to monitor performance of some machines. I understand that the sar command can isolate particular values for more detailed criteria such as Disk I/O. The goal is to provide Disk I/O in a pretty line graph for management and power users to review in an upcoming meeting. The issue I'm having is while attempting to interpret the resulting graph data within Excel. Here's what I've done thus far:

I've told sysstat to log the max history which is 30 days. Then, I use the following command to provide output of said disk I/O from the first day:
Code:
```
sar -b -f /var/log/sa/sa01
```

From here, I use sadf to output to Excel-friendly format:

Code:

sadf -d /var/log/sa/sa01 -- -b | sed 's/;/,/g' > storage01_diskio_1-3-17.csv

For the curious, here's what the output looks like, raw:

Code:

storage01,599,2017-01-01 06:10:01 UTC,1.39,0.00,1.39,0.00,26.92
storage01,600,2017-01-01 06:20:01 UTC,1.32,0.00,1.32,0.00,25.02
storage01,600,2017-01-01 06:30:01 UTC,1.32,0.00,1.32,0.00,24.57
storage01,600,2017-01-01 06:40:01 UTC,1.51,0.05,1.46,1.63,28.13
storage01,600,2017-01-01 06:50:01 UTC,1.41,0.00,1.41,0.00,26.20
storage01,600,2017-01-01 07:00:01 UTC,1.36,0.00,1.36,0.00,25.29
storage01,600,2017-01-01 07:10:01 UTC,1.37,0.00,1.37,0.00,26.09
storage01,600,2017-01-01 07:20:01 UTC,1.33,0.01,1.33,0.04,24.67
storage01,600,2017-01-01 07:30:01 UTC,1.37,0.00,1.37,0.00,25.04
storage01,600,2017-01-01 07:40:01 UTC,1.42,0.00,1.42,0.00,25.33
storage01,600,2017-01-01 07:50:01 UTC,1.32,0.00,1.32,0.00,24.48
storage01,600,2017-01-01 08:00:01 UTC,1.37,0.00,1.37,0.00,25.00
storage01,600,2017-01-01 08:10:01 UTC,1.39,0.00,1.39,0.00,25.19
storage01,600,2017-01-01 08:20:01 UTC,1.39,0.00,1.39,0.00,24.84
storage01,600,2017-01-01 08:30:01 UTC,1.36,0.00,1.36,0.00,25.16
storage01,600,2017-01-01 08:40:01 UTC,1.41,0.00,1.41,0.00,25.65
storage01,599,2017-01-01 08:50:01 UTC,1.33,0.00,1.33,0.00,24.62
storage01,600,2017-01-01 09:00:01 UTC,1.36,0.00,1.36,0.00,24.68
storage01,600,2017-01-01 09:10:01 UTC,1.52,0.01,1.51,0.09,27.80
storage01,600,2017-01-01 09:20:01 UTC,1.34,0.01,1.33,0.04,24.44
storage01,600,2017-01-01 09:30:01 UTC,1.41,0.00,1.41,0.00,25.35
storage01,600,2017-01-01 09:40:01 UTC,1.37,0.00,1.37,0.00,25.07
storage01,600,2017-01-01 09:50:01 UTC,1.34,0.00,1.34,0.00,24.19
storage01,600,2017-01-01 10:00:01 UTC,1.38,0.00,1.38,0.00,24.65
storage01,600,2017-01-01 10:10:01 UTC,605.99,554.65,51.34,6704.93,20238.52
storage01,600,2017-01-01 10:20:01 UTC,1.43,0.00,1.43,0.00,25.04
storage01,600,2017-01-01 10:30:01 UTC,21.78,19.00,2.78,187.34,128.99
storage01,600,2017-01-01 10:40:01 UTC,41.15,21.98,19.17,175.94,221.15
storage01,600,2017-01-01 10:50:01 UTC,1.50,0.00,1.50,0.04,26.61
storage01,600,2017-01-01 11:00:01 UTC,1.48,0.00,1.48,0.00,25.67
storage01,600,2017-01-01 11:10:01 UTC,1.49,0.00,1.49,0.00,26.84
storage01,600,2017-01-01 11:20:01 UTC,1.38,0.00,1.38,0.00,24.81
storage01,600,2017-01-01 11:30:01 UTC,1.43,0.00,1.43,0.00,25.07
storage01,600,2017-01-01 11:40:01 UTC,1.43,0.00,1.43,0.00,25.25
storage01,600,2017-01-01 11:50:01 UTC,1.39,0.00,1.39,0.00,25.39
storage01,600,2017-01-01 12:00:01 UTC,1.40,0.00,1.40,0.00,24.80
storage01,600,2017-01-01 12:10:01 UTC,1.46,0.00,1.46,0.00,26.41
storage01,599,2017-01-01 12:20:01 UTC,1.38,0.00,1.38,0.00,25.30
storage01,600,2017-01-01 12:30:01 UTC,1.43,0.00,1.43,0.00,25.80
storage01,600,2017-01-01 12:40:01 UTC,1.42,0.00,1.42,0.00,26.21
storage01,600,2017-01-01 12:50:01 UTC,1.37,0.00,1.37,0.00,24.67
storage01,600,2017-01-01 13:00:01 UTC,1.42,0.00,1.42,0.00,25.52
storage01,600,2017-01-01 13:10:01 UTC,1.45,0.00,1.45,0.00,25.95
storage01,600,2017-01-01 13:20:01 UTC,1.40,0.00,1.40,0.00,25.46
storage01,600,2017-01-01 13:30:01 UTC,1.39,0.00,1.39,0.00,25.09
storage01,600,2017-01-01 13:40:01 UTC,1.36,0.00,1.36,0.00,24.73
storage01,600,2017-01-01 13:50:01 UTC,1.39,0.00,1.39,0.00,25.68
storage01,600,2017-01-01 14:00:01 UTC,1.37,0.00,1.37,0.00,23.99
storage01,600,2017-01-01 14:10:01 UTC,1.48,0.00,1.48,0.00,27.28
storage01,600,2017-01-01 14:20:01 UTC,1.44,0.00,1.44,0.00,25.65
storage01,600,2017-01-01 14:30:01 UTC,1.44,0.00,1.44,0.00,25.32
storage01,600,2017-01-01 14:40:01 UTC,1.44,0.00,1.44,0.00,25.64
storage01,600,2017-01-01 14:50:02 UTC,1.40,0.00,1.40,0.00,24.97
storage01,599,2017-01-01 15:00:01 UTC,1.50,0.00,1.50,0.00,26.81
storage01,600,2017-01-01 15:10:01 UTC,1.43,0.00,1.43,0.00,26.32
storage01,600,2017-01-01 15:20:01 UTC,1.42,0.00,1.42,0.00,25.36
storage01,600,2017-01-01 15:30:01 UTC,1.44,0.00,1.44,0.00,25.60
storage01,600,2017-01-01 15:40:01 UTC,1.42,0.00,1.42,0.00,25.11
storage01,600,2017-01-01 15:50:01 UTC,1.39,0.00,1.39,0.00,24.71
storage01,600,2017-01-01 16:00:01 UTC,1.39,0.00,1.39,0.00,25.01
storage01,600,2017-01-01 16:10:01 UTC,1.43,0.00,1.43,0.00,25.93
storage01,600,2017-01-01 16:20:01 UTC,1.41,0.00,1.41,0.00,25.40
storage01,600,2017-01-01 16:30:01 UTC,1.39,0.00,1.39,0.00,25.35
storage01,600,2017-01-01 16:40:01 UTC,1.37,0.00,1.37,0.00,24.69
storage01,600,2017-01-01 16:50:01 UTC,1.39,0.00,1.39,0.00,25.75
storage01,600,2017-01-01 17:00:01 UTC,1.37,0.00,1.37,0.00,24.79
storage01,600,2017-01-01 17:10:02 UTC,1.50,0.00,1.50,0.00,27.30
storage01,599,2017-01-01 17:20:01 UTC,1.37,0.00,1.37,0.00,24.56
storage01,600,2017-01-01 17:30:01 UTC,1.39,0.00,1.39,0.00,25.05
storage01,600,2017-01-01 17:40:01 UTC,1.39,0.00,1.39,0.00,24.97
storage01,600,2017-01-01 17:50:01 UTC,1.33,0.00,1.33,0.00,24.20
storage01,600,2017-01-01 18:00:01 UTC,1.39,0.00,1.39,0.00,25.03
storage01,600,2017-01-01 18:10:01 UTC,1.40,0.00,1.40,0.00,25.63
storage01,600,2017-01-01 18:20:01 UTC,1.46,0.00,1.46,0.00,25.89
storage01,600,2017-01-01 18:30:01 UTC,1.39,0.00,1.39,0.00,25.31
storage01,600,2017-01-01 18:40:01 UTC,1.39,0.00,1.39,0.00,24.83
storage01,600,2017-01-01 18:50:01 UTC,1.29,0.00,1.29,0.00,23.96
storage01,600,2017-01-01 19:00:01 UTC,1.33,0.00,1.33,0.00,24.56
storage01,600,2017-01-01 19:10:01 UTC,1.35,0.00,1.35,0.00,25.29
storage01,600,2017-01-01 19:20:01 UTC,1.47,0.00,1.47,0.00,25.92
storage01,600,2017-01-01 19:30:01 UTC,1.37,0.00,1.37,0.00,24.84
storage01,600,2017-01-01 19:40:01 UTC,1.32,0.00,1.32,0.00,24.12
storage01,599,2017-01-01 19:50:01 UTC,1.32,0.00,1.32,0.00,24.29
storage01,600,2017-01-01 20:00:01 UTC,1.33,0.00,1.33,0.00,24.56
storage01,600,2017-01-01 20:10:01 UTC,1.39,0.00,1.39,0.00,26.01

In Excel, I import the data to graph by copying/pasting the output from the previous command. From here, I click the Data ribbon and choose "Text to Columns".
From the new window that appears, I choose the "Delimited" radio button and click Next
I choose the "Comma" checkbox and click Next
"Column Data Format" should be set to General so I click Finish
I now insert a new line above it all for the following headers:
Hostname | Interval | Time | TPS | Reads/sec | Writes/sec | bread | bwrit
While still in Excel, I click the Charts ribbon, highlight all data and choose Scatter->Smooth Marked Scatter.
Lastly, I choose Layout 1 from "Chart Quick Layouts"

As I understand it, the X axis contains time in seconds (or is this milliseconds?), but I'm not exactly sure what the Y axis defines. My output shows 0 20 40 ... in intervals of 20 up to 160. Is this also seconds? I realize I can use gnuplot, but management much prefers Excel charts so I'm making do with what I have. Also, if you do this differently, please share your process in case it can be done more efficiently as I'm sure I'm going about this the long way. Thanks for any/all feedback.

loadedmind · 01-12-2017, 09:48 AM

Huh... No one, eh?

rtmistler · 01-12-2017, 10:05 AM

Well, been subscribed all along but had nothing to say.

I take CSV data and plot it all day long either manually as you say, and then unfortunately using Windows charting utilities in C# because that's the preferred platform for when we plot archived or real-time data to a graphical window.

What you're doing is very normal. Grab the data and plot it. Excel does this fine. Honestly I've never played with any of the Linux alternatives for plotting.

In our cases, a lot of it is biometric sensor data. The Y-axis scalar ranges matter less than the trends which are shown. I.e. an ECG that doesn't look like an ECG is either something wrong with the signal or the patient. The X-axis is very typically time in seconds.

I use 2D line plots and do not try to put any fancy dots, or labeling on them. Further, when you have a mountain of data, for instance 200 Hz of data, changing the representation of the line to be dotted, does absolutely zip because the points are so close together. This however is not a problem in your case if you choose to do this.

Those are my thoughts.

As far as your resolution, I think those time stamps are 10 minutes apart ... 6:10:01, 6:20:01, ... why do you think differently?

loadedmind · 01-12-2017, 11:28 AM

Here's a screenshot to help illustrate. I agree, btw. The raw data suggests 10-minute increments. It's the bottom axis (Y) that I'm questioning.
http://linuxslacker.com