Flagging Processes larger than X

[kfr]

We run multiple SUN Ray Servers and from time to time, some processes tend to grow large over time.

I'm new to zabbix and i'm wondering if it's possible to flag processes larger than x pages.

EG: The ps/nawk command shows the process greater than 300,000k

ps -aef -o vsz,pid,user,args| nawk 'BEGIN {}{if ($1 > 300000) {printf ("%s %s %s %s\n",$1,$2,$3,$4); }}END{}'
VSZ PID USER COMMAND
314744 3981 kfr /usr/local/firefox/firefox-bin

[cstackpole]

While I don't have any experiance with your question directly, I will say this: If you can see an output, then Zabbix can be used to monitor it.

You can write a script to check every x time interval and then send the results to Zabbix via a trapper.

Or you can set up a User Parameter (take a look in the manual) and configure a specific item.

Hope that helps you get started!
Have fun!

[kfr]

cstackpole,

Thanks for pointing me in the right direction.

I setup
UserParameter=chk.vsz,ps -aef -o vsz,pid,user,args| nawk 'BEGIN {}{if ($1 > 300000) {printf ("%s %s %s %s\n",$1,$2,$3,$4); }}END{}'|grep -v VSZ|sort -n

I was able to test with:
zabbix_get -s testsrv -k chk.vsz
304800 9888 kfr /usr/local/firefox-3.0rc1/firefox-bin
383140 2730 brentmc /usr/openwin/bin/Xsun
407032 6737 daniel /usr/local/firefox-3.0rc1/firefox-bin

I setup the item
Name: Processes Using a Large Amount of Memory
Key: chk.vsz
Type of information: Text
Applications: Processes

I just need to figure out the proper trigger (N is confusing me) and then setup an action and I'll be all done!

[cstackpole]

Sorry for the long delay; I was on vacation.

Did you get this to work? Are you having any problems with it?

[kfr]

Sorry for the long delay; I was on vacation.

Did you get this to work? Are you having any problems with it?

I can't seem to get the proper trigger syntax. I have read the docs but i'm confused as to which syntax to use.

I've defined the field as text since it's sometimes multiple lines. eg:
304800 9888 kfr /usr/local/firefox-3.0rc1/firefox-bin
383140 2730 brentmc /usr/openwin/bin/Xsun
407032 6737 daniel /usr/local/firefox-3.0rc1/firefox-bin

[cstackpole]

Hrrm. Yeah I think I see what you mean.

I personally would test out the .diff option for the trigger. Kind of like how the hostname is configured in the templates. The more I think about it, I would actually configure it just like that and see how it works. If I understand you right in what you want to do, that should work.

Have fun!

[kfr]

I managed to get the trigger working like you suggested.

I have another question though.

I setup a new userparameter to monitor system temperature

The output of the userparameter is:
mb_p0_t_core 58
mb_t_enc 22

The above shows the processor 0 temperature as 57 celcius and the motherboard temperature as 22. It's possible that, on some servers, there are more than 1 cpu.

I'm able to see a "History" of the userparameter if I set the type of information as text.

How can I get that data into a single graph, with separate entries for each processor(s) and motherboard temperature sensor?

[tighep]

I would probably modify the user paramater to take a variable for each processor, or the motherboard and return just the value of that particular temperature. This will allow the type not to be text, and will make monitoring and graphing each individual piece easier. I have done a similar task for monitoring our tape drives with Tivoli Storage Manager, and created individual items for each drive, this allows me to easily monitor each drive individually.