Help with colletcl tool

peter79

Hey Guys,

Im fairly new to the world of linux so I need a bit of help. Im currently using a tool called collectl (http://collectl.sourceforge.net/) to gather sone I/O information about a Red Hat server. Ive figured how to capture the data but my problem is I really need to get the results into a CSV file for Excel.

Has anyone on this forum any experience using collectl and if so can the data be exported to a CSV file?

Thanks.

markseger

I'm actually the author of collectl! Personally I don't like csv formatted data because I find it a pain to parse and most utilities that read CSV also read space separated data so that's the default when you with with -P. If if you include -f it ill write to a file in that directory. BUT if you really want csv data , you can use "--sep ','" to tell it to use a comma as the separator instead. Does that answer your question or are you looking for more details?
-mark

peter79

Mark,

Thanks for your reply. Where in the command would you use --sep? For example if I ws running

colletl -sd -c10 --P -p/temp

Also there is a chance I might have to gather the data for up to a week, how big do you think the csv file would get?

Thanks in advance.

By the way its a great tool!

markseger

If you really only care about disk data and nothing more - it's not that expensive to grab more - the question is do you just want summaries of totals across ALL disks or would you prefer data on individual ones? -sd will get you the summary, -sD the details and -sdD both. Anyhow assuming you really want summary as you've indicated above AND if you want it for 5 days, run collectl as a daemon. just edit the line DaemonCommands in /etc/collectl.conf from:

DaemonCommands = -f /var/log/collectl -r00:00,7 -m -F60 -s+YZ

to

DaemonCommands = -f /var/log/collectl -r00:00,7 -sd -P --sep ','

that will write the csv data to a file in /var/log/collectl, creating a new one each midnight. To start it just do /etc/init.d/collectl start and after 5 days /etc/init.d/collectl stop. The default collection interval is 10 seconds.

The output, which you can generate interactively to look at would be from the command:

"collectl -sd -P --sep ','" and look like the following, noting the interactive default interval is 1 sec:

[root@poker collectl]# collectl -sd -P --sep ','
waiting for 1 second sample...
#Date,Time,[DSK]ReadTot,[DSK]WriteTot,[DSK]OpsTot,[DSK]ReadKBTot,[DSK]WriteKBTot,[DSK]KbTot,[DSK]ReadMrgTot,[DSK]WriteMrgTot,[DSK]MrgTot
20110111,08:04:36,0,0,0,0,0,0,0,0,0
20110111,08:04:37,0,2,2,0,264,264,0,64,64
20110111,08:04:38,0,0,0,0,0,0,0,0,0

this should be enough to get you going...
-mark

peter79

Mark,

Thanks again for yout response just to give some background info. The data I need to captutre is wtites per second on the disk subsystem. So I will be using -sd. I would like to capture the at least 3 days worth of data mayne up to 5.

I would like to have the data captured as a CSV file so it can be viewed in excel.

I tried running the command below because I want to see what the results look like in Excel. However it did not run is the syntax correct?

collectl -sd -P -p/temp --sep','

markseger

why didn't you say excel?
excel can read collectl's space-separated format with no problems, you just need to open as you would any file and tell excel the fields are space-separated when it asks.

Anyhow if you want to capture disk writes/sec, -sd will show the total writes on all your disks. If all but one are idle, this should work fine, but if you have multiple disks active, -sD will get you exact numbers for each.

I don't know where -p in your command above is coming from. You need -f/temp if you want to write the output to the directory /temp. By default, if you have the compression library installed, collectl will write its output in a compressed 'gz' file, which you can always manually uncompress. Alternatively you can add the switch -oz which will tell it not to compress.

So, the command I'd consider using is:

collectl -sD -P -f/temp -oZ

then try importing the /temp/*dsk into excel without around with --sep. It's really not hard...
of you use -sd you'll have a file named *.tab instead. if you use -sdD you'll get both!

-mark

peter79

Mark,

I really appreciate all yout help on this. I have tried opening the tab output files in Excel and while it does open the file it just was a bit of a mess (I have attached the excel file tp this post).

Is it possbile to to format the output so that only the writes pe second is recorded? Also I tried configuring collectl to run as daemon but when I attempted to start the service it failed. Please note that I'm currently testing this out on a ESX server (ESX is built on a red hat kernel so it shouldnt be a problem.