2

I search a generic hardware bottleneck detector.

The tool should monitor all processes on the server for 24 hours.

Then it should report where the bottleneck could be.

Required features:

  • open source
  • works for linux servers

List of possible bottlenecks:

  • Amount of CPU
  • Single CPU speed
  • IO performance
  • Network
  • ...

A tool which just shows a graph of cpu/io usage does not solve my question.

The question is: where does the most hardware related waiting occur?

Improve this question

1 Answer 1

Reset to default
1

A robust solution would be to use collectd with an appropriate frontend, for example Grafana. This requires a working a database setup.

A simpler solution would be to use icinga2. This requires you to at least write or download a script/plugin for network monitoring, CPU/Memory/IO scripts are included in the default installation.

On many linux distributions there are packages named along the lines of monitoring-plugins or nagios-plugins which include more useful monitoring scripts.

I personally use both methods in conjunction, icinga for reporting problems and Grafana for analyzing them afterwards, though Grafana is theoretically able to do both.

Improve this answer
4
  • 1
    Yes, the tools you mention could solve this. But finally there is still a lot of human expert knowledge needed to understand the data. If I want to, and have time, then I can analyse the data and draw a conclusion. But I am lazy :-) With other words: Your answer is a good start. It shows how to fetch data. I search for an automated analysis of the data.
    – guettli
    Dec 6, 2018 at 9:50
  • Your question seems underspecified then. Icinga an grafana can send you a report mail/sms whatever if a defined scenario occurs i.e. 10min of >80% load. What exactly are you looking for then?
    – Sheppy
    Dec 6, 2018 at 10:00
  • I updated the question. See the last two sentences of it. Is it more clear now? If not, please tell me.
    – guettli
    Dec 7, 2018 at 12:15
  • Well the most waiting is likely to be IO blocking, mostly disk that will be, a overloaded network will usually result in requests being dropped, as will high CPU load (for webservers anyway), even with the updated question icinga (or nagios or any other mainstream monitoring application) with plugins or collectd with grafana and mail notifications is the way to go you can easily specify circumstance under which you will receive a notification, which still seems to be exactly what you are looking for
    – Sheppy
    Dec 9, 2018 at 10:53

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge that you have read and understand our privacy policy and code of conduct.

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