Operating System - HP-UX
1834869 Members
2533 Online
110070 Solutions
New Discussion

Re: Network device utilization

 
SOLVED
Go to solution
Wagner Villela
Advisor

Network device utilization


Dear friends

I need to know which process is more using a network device in determined moment (lan2 or lan3). How i can make this? I use hp-ux 10.20.

Thank you

Wagner
Wagner
5 REPLIES 5
Helen French
Honored Contributor

Re: Network device utilization

I would use GlancePlus -> Network for getting this information. You will get all information per interface (packets in, packets out, errors, load etc) fro the menus:

# gpm

Another command would be:

# netstat (see man pages for details)
Life is a promise, fulfill it!
sven verhaegen
Respected Contributor
Solution

Re: Network device utilization

Hi

what you are doing is very hard to accomplish , the data from Glance is normally information per interface or information per process , I have no knowledge of information linking the interface to the processes themselfs , that is because in fact there is no correlation between a lancard and a process , it is the routing stack which directs the process request to the appropriate port of the system , I doubt if even perfview is able to make that kind of asessement , Glance is indeed your best shot but as said I'm not aware of a function able to tdo this in Glance
...knowing one ignores a greath many things is the first step to wisdom...
Wagner Villela
Advisor

Re: Network device utilization

Sven

I think you is right.

Through the Glance is not possible to identify which process cause my switch freezing. We have a broadcast storm in determined moments.

The maximum that i obtain is to see through glance the incredible increase in the number of packages passing through for a network device. I don??t obtain to identify the responsible process therefore.

Thanks

Wagner
Wagner
Stefan Farrelly
Honored Contributor

Re: Network device utilization

The graphical version of glance, gpm, will show this for you - in a roundabout way.

Run gpm, bring up process list, add metrics non disc logical reads and non disk logical writes. Sort columns so processes are listed by these 2 metrics (and you'll probably have to arrange columns to show these 2 first, then pid and user). Drop measurement interval to 5 seconds or less, then when a user does a large ftp (for example) either reading or pushing, it will show up under these 2 metrics above. Basically it sorts processes by network usage. My test doing large ftp's showed up easily on the above example - I could see which user was doing them and if sending or receiving ftp files.

This should help you find the process(es) hitting your network hard.
Im from Palmerston North, New Zealand, but somehow ended up in London...
Lukas Grijander
Advisor

Re: Network device utilization

Hi.

whith lsof you can list all processes "attached" to an ip address :

lsof -i @192.168.1.1

Hope it helps.
Rafa