Re: [B.A.T.M.A.N.] Recover in case of node fail

[Sven Eckelmann <sven@narfation.org>]

[-- Attachment #1: Type: text/plain, Size: 1543 bytes --]

On Wednesday 22 July 2015 15:26:23 Carlos Meralto wrote:
> Hi guys,
> 
> I'm trying to make a study of various protocols for WMN and one of
> them is BATMAN. One the tests is the time to recovery in case of node
> fail.
> 
> I have a topology with 4 routers with 2 possible paths:
> A-B-C
> A-D-C
> 
> For the test i'm doing a continuous ping from A to C (level 3 ping)
> and shut down the router that forward the message (i.e B or C). In
> this test BATMAN can't recover the PING. But when i cancel the ping
> and check the Originators table (batctl o) the route is refreshed. And
> when i try to ping again it works.
> 
> Is any limitation in BATMAN for recovery in a continuous PING?

No, sounds more like you are not using batman-adv. Because
batman-adv doesn't know about the l3 stuff. It just
transfers ethernet frames from A to C. B and D are
chosen automatically for each frame (and not for a
l3/l4/... connection).

A `batctl o` does not trigger any route re-calculation.

You can use batctl ping with -R to see the hops and the
switch of the route when D fails.

Btw. if you have a setup with many dynamic changes then you
may want to reduce the originator interval on each node.

> Has anyone done this type of test and values obtained?

Yes, there is a long list of papers doing such things with
batman-adv, bmx, olsrd, babel, ...

And the fall-over worked fine in my last tests. There is
always a small time before the disappeared node is detected
by the tq metric but it is nothing breaks normal connections.

Kind regards,
	Sven

[-- Attachment #2: This is a digitally signed message part. --]
[-- Type: application/pgp-signature, Size: 819 bytes --]