NRPE vs SNMP monitoring?

[dinush]

I used to monitor through nrpe agent and for easier agent deployement, we changed to agentless SNMP monitoring. Is SNMP monitoring is a better option? What are the advantages and disadvantages?

[rkennedy]

SNMP has a limited set of capabilities, where as NRPE will allow you to have a full range of execution available.

The general SNMP doc we have is here - https://assets.nagios.com/downloads/nag ... g_SNMP.pdf

The other option that is up and coming is NCPA - https://www.nagios.org/ncpa/ - which will be similar to NRPE, but have quite a few more capabilities.

[dinush]

i moved to snmp recently. That means my capabilities are limited now. Thought SNMP is better.

I did change because nrpe needed to compile/install in every hosts , but snmp need only 2-3 lines to be added to the hosts with net-snmp rpm installed.

Please advice me what should i do now. Back to NRPE ?

[rkennedy]

I enjoy SNMP for the ease of use as well.

The question at this point, becomes what are you after monitoring that you cannot with SNMP currently? There may be a way to monitor it without the agent, but we'll need to know more.

[dinush]

we need to monitor and get graphs on below stats..

CPU user per CPU core
system Per CPU core
iowait Per CPU core
idle Per CPU core

Host load average Per host

Memory used Per Host
cached Per Host
real Per Host
total Per Host
swap used Per Host
swap total Per Host

Disk IO writes/s (KB) Per disk partition
read/s (KB) Per disk partition

Network IO RX bytes /s Per NIC
TX bytes /s Per NIC

TCP connections ETABLISHED Per host
TIME_WAIT Per host
FIN_WAIT Per host
CLOSE_WAIT Per host

[rkennedy]

Going to scatter my thoughts to your different sections -

Code: Select all

CPU	user	per CPU core
system	Per CPU core
iowait	Per CPU core
idle Per CPU core

I believe the CPU usage per core is usually available for SNMP, I am not sure about the remainder though.

Code: Select all

Memory	used	Per Host
cached Per Host
real	Per Host
total	Per Host
swap used	Per Host
swap total	Per Host

The below link mentions the OID's for most of what you're after here for memory, all should be available.
http://www.debianadmin.com/linux-snmp-o ... stics.html

Code: Select all

Disk IO	writes/s (KB)	Per disk partition
read/s (KB)	Per disk partition

Not sure if this is readily available, would need to take a look at the walk.

Code: Select all

Network IO	RX bytes /s	Per NIC
TX bytes /s	Per NIC

This should be possible based on taking the amount of data provided and comparing to another walk taken a second before. I believe there are plugins that do so. (just give you a general answer, as I don't know off the top of my head. might be worth making a new post asking about which plugin for SNMP network monitoring.)

TCP connections ETABLISHED Per host
TIME_WAIT Per host
FIN_WAIT Per host
CLOSE_WAIT Per host

As the walk generally shows all of this information, I believe this would be possible as well. You would just need to count the amount of connections per state type to get the metric.

For example -

Code: Select all

TCP-MIB::tcpConnState.x.x.x = INTEGER: listen(2)
TCP-MIB::tcpConnState.x.x.x = INTEGER: established(5)

[tmcdonald]

Just checking in since we have not heard from you in a while. Did @rkennedy's post clear things up?