Hello,
we are using MRTG to measure throughput connections between LAN equipment, and bbmrtg.pl to report the status to Hobbit / Xymon. bbmrtg.pl is extended to optionally check the status of a connection too, and report a 'red' status if the connection is down. We have found this a useful feature, as it detects link problems even in case redundancy in the network hides the problem.
bbmrtg.pl is extended to take some special action if (A) variable bb*chklink is defined and (B) at the same time no input is received, thus 0 bytes per second, and almost no output is sent. There are two cases: if bb*chklink is negative, this condition is sufficient to set the status of the connection to 'red'. If bb*chklink is positive, it is assumed to be the ifIndex of the interface. Using SNMP, the current status of the interface is retrieved. If the status is "down", the connection status is set to 'red'. The small amount of output allowed takes care of some L2 protocols which are almost always active, such as STP or CDP.
Note that this condition cannot be checked with the original version of bbmrtg.pl. That version cannot check for the condition that both input *and* output is (almost) zero.
Attached is the list of modifications in bbmrtg.pl.
Kind regards, Wim Nelis.
The NLR disclaimer (http://www.nlr.nl/emaildisclaimer) is valid for NLR e-mail messages.
On Tue, 2008-11-18 at 14:54 +0100, W.J.M. Nelis wrote:
Hello,
Attached is the list of modifications in bbmrtg.pl.
that's a good start, but it doesn't take a few things into consideration:
- what if snmp V3 is in use instead of snmp V2?
- What if each router uses a different snmp community string?
- how do you score it if there are more than one interfaces associated with a bb*host?
I think you are going to need a bit more logic to make it generally useful.
-- Daniel J McDonald, CCIE #2495, CISSP #78281, CNX Austin Energy http://www.austinenergy.com
Attached is the list of modifications in bbmrtg.pl.
that's a good start, but it doesn't take a few things into consideration:
- what if snmp V3 is in use instead of snmp V2?
- What if each router uses a different snmp community string?
In both cases, one can use the condition (input == 0 && output == small) as sufficient to signal an error condition, thus skipping the SNMP (double) check. In my experience, this gives hardly any false positives. In our network, about 300 connections are checked in this way, of which about 100 are configured without an SNMP check.
- how do you score it if there are more than one interfaces associated with a bb*host?
We have collected all the inter-switch links (trunks) on one switch in one test, called "trunk". It shows one graph, including is status green, yellow or red, per trunk on that switch. The failing trunks are shown at the top of the page. The status of the test <switch>.trunk becomes the worst status of all trunks on that page.
I think you are going to need a bit more logic to make it generally useful.
I am wondering what you mean which "generally". This modification is intended to be used in specific situations. In our case redundancy build in the network may mask the failure of one connection (trunk). The users won't notice the failure, but the network administrator should know about it. On the other hand, I think the extension to bbmrtg.pl is generally usable for those who have this specific need.
Wim Nelis.
The NLR disclaimer (http://www.nlr.nl/emaildisclaimer) is valid for NLR e-mail messages.
On Wednesday 19 November 2008 10:25:08 W.J.M. Nelis wrote:
I am wondering what you mean which "generally". This modification is intended to be used in specific situations. In our case redundancy build in the network may mask the failure of one connection (trunk). The users won't notice the failure, but the network administrator should know about it. On the other hand, I think the extension to bbmrtg.pl is generally usable for those who have this specific need.
Devmon does all this for me without any hacks (as it has separate tests for the interface status and interface load) ... and the new devmon weathermap script will actually show any links down as red ...
Buchan Milne <mailto:bgmilne at staff.telkomsa.net> wrote:
Devmon does all this for me without any hacks (as it has separate tests for the interface status and interface load) ... and the new devmon weathermap script will actually show any links down as red ...
By what method(s) is devmon detecting that a link is 'down'? I presume you, er, hacked up some devmon rules to do this? ;) Maybe you can share these with us? I've not used devmon, so forgive me if they are elementary - it would still be nice to see them or the config needed to get this working, and it might encourage more people to use it and your nice looking devmon-weathermap. Or if this link detection is completely part of devmon by default, then can you explain how devmon concludes a link is up or down?
Many thanks,
SebA
participants (4)
-
bgmilne＠staff.telkomsa.net
-
Dan.McDonald＠austinenergy.com
-
nelis＠nlr.nl
-
spa＠syntec.co.uk