[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]
Re: [hobbit] Snmpcollect in the current snapshot
- To: hobbit (at) hswn.dk
- Subject: Re: [hobbit] Snmpcollect in the current snapshot
- From: henrik (at) hswn.dk (Henrik Stoerner)
- Date: Mon, 10 Sep 2007 22:52:04 +0200
- References: <796B35279263C14D8E0E2D9137434D7B34C4E5 (at) tango.Revere.local>
- User-agent: Mutt/1.5.13 (2006-08-11)
On Mon, Sep 10, 2007 at 11:56:55AM -0700, Levin, Alexandre wrote:
> In the current snapshot I'm glad to see new file hobbit_snmpcollect, is
> it a replacement for devmon?
No, you should still use devmon for monitoring SNMP enabled devices.
At least for the time being - devmon, I think, can do a lot more than
snmpcollect right now.
> Any tips on usage?
> Or it is too early to try it ?
It's very early to try this, only do it if you're really adventurous.
I only started this Saturday, so it hasn't seen any real testing.
Nonetheless, I setup a small test run of it this afternoon, and so far
it appears to be working as intended. If you do want to try it out,
get the snapshot version which will be available in a couple of hours.
The reason for doing this is that our network people are having problems
collecting statistics with MRTG. Basically, their MRTG server is
extremely overloaded, and besides that they'd rather not have to worry
about setting up MRTG and maintaining a website for it. So I was asked
if it was possible for Hobbit to handle the ~10.000 datasets they are
currently collecting with MRTG. Of course, that was a challenge I had to
meet :-)
I've chosen to be a bit more thorough than MRTG in the data collection.
MRTG by default collects only the in/out byte counts from a device; you
can do that with Hobbit's new snmpcollect, but a better choice is to
grab the entire set of data present in the "interfaces" MIB. This
includes some interesting data, e.g. how many bad packets arrived over
the interface, how many were discarded, and what the current output
queue length is.
snmpcollect uses a configuration file that looks like this:
[localhost]
version=2
community=public
mrtg=2 IF-MIB::ifInOctets.2 IF-MIB::ifOutOctets.2 Main IO
# ifmib=2 LAN
# ifmib=(eth0) LAN
# ifmib=[0:e:a6:ce:d6:85] LAN
ifmib={172.16.10.100} LAN
So for a host, you define the SNMP version you want to use (currently v1
and v2c only), the community name, and optionally an IP address for it.
Then you can request snmpcollect to fetch either a set of data in MRTG
form, or you can fetch the entire interface-MIB set of data for a router
interface / switch port. There are different ways to define what
interface you want the data for - by interface number, description, MAC
address or IP. snmpcollect grabs the data and sends it to Hobbit in a
"data" message; this is picked up by the current hobbitd_rrd module and
put into trend graphs.
Right now it just collects the data. I have some ideas to add monitoring
so it might trigger an alert if the error-rate of an interface goes up,
or the traffic exceeds certain limits (min/max). Nothing definite yet.
Oh - and the current snapshot blindly assumes you have Net-SNMP
installed; if not, it won't compile. No need to tell me that - I know.
Regards,
Henrik