[NCSA-discuss] MIBs and monitoring

[Steve Wolfe]

One comment:
I was able to get good metrics out of net-snmp for servers

They have added a server MIB that has many good things like cpu
utilization, processes, disk utilization

but
MRTG seemed to be the level of complexity I wanted on the screen, but
couching the oid calls in terms that MRTG liked was brutal and often
error prone.

These looked daunting but maybe good.
RRD and  cacti

Steve

On Tue, Oct 11, 2011 at 11:00 AM, Craig Cook <cncook001 at yahoo.com> wrote:
> SNMP is a universal way of reporting metrics.  If you want to report to most available monitoring systems that is the way to go.  It would be much easier for the LVS project to simply support SNMP.
>
> If someone in the xymon/nagaios/whatever community really did not like using snmp they would then have to support their native way of getting data.
>
>
> When you poll the native Solaris SNMP agent it contacts the kernel to get metrics.  I am not sure how it does it though.
>
> LVS SNMP may run admin utils and returns metrics or it may have direct calls somehow.
>
> When you poll large numbers of SNMP devices it adds load on your central polling server.  You can't have a single threaded app doing the polling (I've tried ;).
>
>
> I guess it comes back to scale, if you have 100's of LVS devices it may be better to get them to send data to the central server, if you only have a few I would use SNMP to poll them.
>
>
> Craig
>
>
>
> ________________________________
> From: Joseph Mack NA3T <jmack at wm7d.net>
> To: Craig Cook <cncook001 at yahoo.com>; NC*SA Discussion List <ncsa-discussion at ncsysadmin.org>
> Sent: Tuesday, October 11, 2011 10:32 AM
> Subject: Re: [NCSA-discuss] MIBs and monitoring
>
> On Tue, 11 Oct 2011, Craig Cook wrote:
>
>> I don't recommend SNMP monitoring for servers.
>
> OK use utilities unless you can't, in which case you go to snmp.
>
>> Yes, you can extend net-snmp to run scripts and provide metrics when you poll an OID but it becomes difficult to update clients.
>
> I work on the Linux Virtual Server (LVS) project. Someone has wrote an snmp agent for LVS to monitor what is easily monitored by other people with the standard setup and admin tools (which read /proc). I haven't looked at how it works; perhaps it calls the admin utils underneath and returns snmp messages. Why might they have done that? snmp only shop? they like snmp? they have nothing better to do with their time?
>
>> When I was last using MRTG there was a limitation of only being able to have 2 plot lines (which is fine for traffic utilization).
>
> I run rrdtool now which I'm happy with.
>
> thanks
> Joe
>
> -- Joseph Mack NA3T EME(B,D), FM05lw North Carolina
> jmack (at) wm7d (dot) net - azimuthal equidistant map
> generator at http://www.wm7d.net/azproj.shtml
> Homepage http://www.austintek.com/ It's GNU/Linux!
> _______________________________________________
> ncsa-discussion mailing list
> ncsa-discussion at ncsysadmin.org
> http://www.ncsysadmin.org/mailman/listinfo/ncsa-discussion
>