I have been reading about SNMP,I have to set-up some server with centos, debian. The idea is to use some snmp manager for monitoring these servers.

I have read how to install snmp in linux, know what is a mib, OID. I installed net-snmp and net-snmp-utils

my doubt is about how to integrate or monitor the server from the manager/monitor.


1- Do I need to add the mibs from my linux server to the snmp manager? or do I just need add the IP and the community.

what I look for monitor is cpu usage, memory, disk space, and some processes of an application that would be running in the server.

2- how frequently the monitor poll the devices to know if there is something wrong.

3- I would like that is there is any error, the agent send a trap to the server to notify about this event. what configuration does it require in the manager/agent side?


thanks

Mike:
tell me more about this host.

Subscribed with interest...

The word "doubt" and the word "question" mean different things.
You load the MIB on the server to be monitored. From there, you can then run an SNMP query which will return the new set of values that the MIB represents.
Those are all pretty much there in the 'standard' SNMP configuration, in the ".enterprise" or "system" classes. Read the man page on snmpwalk, or see some of the THOUSANDS of easily-found examples on Google.
You're the administrator, so you decide how often you want to poll things.
When you set up SNMP on the server to be monitored, you can also set up where to send traps, and what to send. Again, read the man page on snmpd.conf:
http://www.debianhelp.co.uk/manpages/snmpdmanpage.txt

If you're doing an interval poll, sending traps won't get you alot, since the system monitor will only be behind by whatever the maximum poll rate is. For example, if the poll rate is 5 minutes, and a service fails, the LONGEST it could be down is < 5 minutes. Sending a trap sooner won't buy you a whole lot.

Nagios and Zabbix are great monitoring tools that can use SNMP, and they are also amply documented.

zabbix is rock solid.

thanks..

The monitor we are using is solarwinds.

I can monitor the disk, cpu, memory. the basic default stuff, at least.

Now I want to monitor some processes, like httpd, oracle and some other 3rd party services.

in the snmpd.conf I can set proc httpd, proc sshd and then query it with snmpwalk with the OID .1.3.6.1.4.1.2021.2

I am trying to configure solarwinds to show the prTable and launch an alert if a process is not running, I cant find this option in the alert section so far. it let me query for a column of the table but not the whole information of the table...

meanwhile I am not sure if the "proc" should be used with the intention of creating an alert and send a trap...or if the manager should detect it automatically, or both...

So really much more of a solarwinds question than SNMP really. Have you tried solarwinds support?