Asymmetric routing:  Host A has more than one interface, say interfaces
1 and 2, and a packet is sent to a destination B via interface 1 and a
response packet from B is received on interface 2 instead of interface
1, this is asymmetric routing.  Extending this to the routing hosts in
the network (routers), asymmetric routing occurs anytime a packet does
not use the same interfaces going in both directions.  There are all
sorts of reasons for this in the network, some legitimate, some
illegitimate and some intentional.  It is really only a problem if the
packets from A to B or B to A fail to actually arrive.  It is also,
generally, and I emphasize, generally, not always, an indicator of
something "not quite right" for an end-point client or server host,
which is why these warning messages are generated.

RIP, a very dumb routing protocol, tends to experience this more than
other protocols which carry more information about network state because
of higher granularity in the communicated metrics.

My responses to Zach are pointing out that he has very limited ability
to actually get the messages to go away and fix the "not quite right"
situation.  And, if the quantities of these messages becomes a problem,
that is a different situation and needs resolution even if nothing in
the network is actually "broke."

Routing (asymmetric or otherwise) is a function of the IP stack
forwarding table (the routing table).  The contents of this table are
determined by interface definitions (IP addresses and subnet masks),
preconfigured routes and routes learned via routing protocols. Routing
protocol deamons learn the routes and "inject" the information into the
IP stack of the host merging all three sources into a single picture of
the network to the host at any given moment in time.  There is
ultimately only one forwarding table.  Some standard rules exist for
route selection, but the IETF RFC's are ambiguous on how exactly
multiple routing protocols should interact with each other in the same
host and this becomes implementation specific.  Cicso does it one way.
A Linux host might do it another.  And a different router manufacturer,
say Juniper, again differently.  But, this digresses.

Harold

Gregg C Levine wrote:

Hello!
I understand the idea behind asymmetrical routing, (Thank you Harold!), but
for the folks who are rather new to the ideas behind how Linux does his
routing, would one of you carefully reiterate the whole concept?

Oh and while I am at it, Zach, would you explain what prompted your
suggestion for the subject?

--
Gregg C Levine [EMAIL PROTECTED]
"The Force will be with you. Always." Obi-Wan Kenobi





-----Original Message-----
From: Linux on 390 Port [mailto:[EMAIL PROTECTED] On Behalf Of Zach
Pratt
Sent: Tuesday, August 07, 2007 5:15 PM
To: [email protected]
Subject: Re: [LINUX-390] occasional martian source messages

I've done some further reading on the topic of asymmetrical routing,
and I'm wondering if I should set up two routing tables (one for each
interface). Any suggestions?




----------------------------------------------------------------------
For LINUX-390 subscribe / signoff / archive access instructions,
send email to [EMAIL PROTECTED] with the message: INFO LINUX-390 or visit
http://www.marist.edu/htbin/wlvindex?LINUX-390




----------------------------------------------------------------------
For LINUX-390 subscribe / signoff / archive access instructions,
send email to [EMAIL PROTECTED] with the message: INFO LINUX-390 or visit
http://www.marist.edu/htbin/wlvindex?LINUX-390

Reply via email to