0

Problem

Sometimes the io on our servers is higher than we want it to be.

I want an automated process which logs all needed information to debug the problem later.

Current strategy

I want to take a snapshot of the current linux processes on my server every ten minutes.

For me it is very important to see the information per process. AFAIK tools like sar don't show me these important details.

Then I want a way to analyze the created log output.

Background information

If you look only at the summary of the total io (like sar does), then it is easy. You can paint nice diagrams with peeks.

But up to now I found no tool which helps me to look at io over time and process.

I want to avoid scripting iotop

I know the tool iotop which has a batch option.

I could do some scripting and help myself. But maybe there is already a nice tool which can help me to find the processes which creates the io peeks.

Which tool could help?

I guess this problem is not new. There are several questions like this at serverfault and stackoverflow.

Most of them get the answer "use iotop".

But iotop is for interactive usage. The batch option exists, but this is only the first step (collecting data).

I guess the second step (analyzing the data) does everybody with magic shell tools up to now.

... unfortunately I am no shell wizard.

I know this question is not easy to solve, since I want both: Aggregation (high level) and drill down of individual processes (single value)...

1

I would suggest taking a look at GRR which is a python framework for doing exactly the sort of things that you are looking to do.

  • Wow, looks great. I must have been blind. Thank you Steve! – guettli May 17 '17 at 8:45

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.