Monitor all process mem/cpu usage over time

[d1nd141]

Hello,

at the moment we are starting the migration process from IBM ITM6 to Zabbix.

ITM has a feature to monitor ALL Processes of Linux&AIX servers with the memory & cpu usage and some other stuff for every single process.
So it's easy to make stats about a single process over time. (so basically a "live" top)

We know it would be possible to inject all this data to splunk or other tools, but we did not find yet a solution to inject to zabbix.
We tested this as custom script send as log, so we have a list of process, but doing stats for a process is not possible.

Anyone an idea?

Thanks

[kloczek]

Such monitoring does not make any sense because most of the processes are running only for very short period of the time.
Just look on current PID and multiply that by memory, cpu metrics and you will receive actual number of metrics which will be necessary to create.

[d1nd141]

In my opinion it makes sense.
F.ex one server has a problem, it will be of interest to see which process may have caused the problem and see the cpu/mem usage over time.

We use this feature in ITM frequently.
Even if the pid is changing, you can filter on process name and see the graph stats.
I know it will take lots of space, f.ex. in ITM datawarehouse we have over 400GB of RR storage. But in case of a problem, it is really useful.