Hi,

The first collector has an IP of 192.168.1.1.
It has a flow-fanout wich was started like this:

flow-fanout 0/0/2054 127.0.0.1/0/2054
192.168.1.1/192.168.1.2/2054

It has a flow-capture listening on 127.0.0.1:2054.

I have another collector at 192.168.1.2. 

Notice the logic of the flow-fanout. It says, for each
flow arriving on all interfaces "including loopback"
as denoted by "0/" in the first flow-fanout argument,
send a copy to flow-capture which is listening on
127.0.0.1/0/2054, and to 192.168.1.2

When the packet arrives at the external interface
(192.168.1.1, a copy is send to its loopback
interface, and with the packet arriving at loopback,
it will be seen again by flow-fanout as a new flow,
having a parameter that tells "send a copy of all
flows arriving on all interfaces", which creates a
possible infinite loop, possibly sending tons of flows
to another remote collector(s). Is there any way that
the flow-fanout will detect that this flow he is
currently seeing on loopback is the same as the one he
saw previously on his external interface (in my case
192.168.1.1).

I guess, the first flow-fanout argument should have
been 192.168.1.0/0/2054, but as I see it(personal
opinion only) , it is a limitation of the program
being unable to anticipate this kind of looping
scenario. Or was it originally by design, that user's
should watch out what they put between these ///'s.


Recently we just had a network wide outage.
If that was what have caused our entire network to
collapsed, eating up almost 100Mbps of bandwidth, then
I guess its a good thing that I have uncovered a
weakness in our LAN infrastructure. Our clients
connected via E1 links could not possibly saturate our
entire network even summing up their entire usage with
the rest of our clients E1 connections. However, hosts
having 100Mbps connection such as our LAN users, are
free to saturate the entire upstream pipe.
It has already happened to my previous job, but the
responsible host is another host, which can easily be
seen by the netflow collector, unfortunately, this
time (although not 100% sure), it was the netflow
collector itself who is probably causing the problem.


Any suggestions?






__________________________________________________
Do You Yahoo!?
Tired of spam?  Yahoo! Mail has the best spam
protection around 
http://mail.yahoo.com 

__________________________________________________
Do You Yahoo!?
Tired of spam?  Yahoo! Mail has the best spam protection around 
http://mail.yahoo.com 
_______________________________________________
Flow-tools mailing list
[EMAIL PROTECTED]
http://mailman.splintered.net/mailman/listinfo/flow-tools

Reply via email to