olsr-users - a mailinglist where users can post questions and discuss olsrd related issues.

John Clark | 26 Jun 2012 03:59

OLSR mesh with loop potential... having problems.

I've not had to look in to OLSRD and 'mesh' problems for quite some time.

However, I have a new project and am having problems.

The version of OLSRD i'm using was one that is part of the Debian package set, and has the following version number:

0.5.6-rc7

I downloaded the most recent src from www.olsr.org, and there seems to be some differences in the config
file, and it was not able to find the
plugins properly.

In any case, on looking for others who have had problems with olsr and looping, it seems this has been a
problem for some time. I don't know why
I didn't see this before when I was using it more, but perhaps most of my 'meshes' were in fact, rather linear
or trees and did not contain loops.

Does anyone have some suggestions about dealing with mesh configurations that have potential loops?

Thanks,
John Clark.

--

-- 
Olsr-users mailing list
Olsr-users <at> lists.olsr.org
https://lists.olsr.org/mailman/listinfo/olsr-users

headers

Hans-Christoph Steiner | 26 Jun 2012 04:30

Re: OLSR mesh with loop potential... having problems.


I think you should try using a newer version of the package, you can
download pre-built ones here, which should also work for Debian:

https://launchpad.net/~guardianproject/+archive/commotion

I'll also look into putting a newer version of the olsrd package in
Squeeze/backports.

.hc

On 06/25/2012 09:59 PM, John Clark wrote:
> I've not had to look in to OLSRD and 'mesh' problems for quite some time.
> 
> However, I have a new project and am having problems.
> 
> The version of OLSRD i'm using was one that is part of the Debian package set, and has the following version number:
> 
> 0.5.6-rc7
> 
> I downloaded the most recent src from www.olsr.org, and there seems to be some differences in the config
file, and it was not able to find the
> plugins properly.
> 
> In any case, on looking for others who have had problems with olsr and looping, it seems this has been a
problem for some time. I don't know why
> I didn't see this before when I was using it more, but perhaps most of my 'meshes' were in fact, rather linear
or trees and did not contain loops.
> 
> Does anyone have some suggestions about dealing with mesh configurations that have potential loops?

(Continue reading)

headers

Henning Rogge | 26 Jun 2012 07:57

Re: OLSR mesh with loop potential... having problems.

On 06/26/2012 03:59 AM, John Clark wrote:
> I've not had to look in to OLSRD and 'mesh' problems for quite some
> time.
>
> However, I have a new project and am having problems.
>
> The version of OLSRD i'm using was one that is part of the Debian
> package set, and has the following version number:
>
> 0.5.6-rc7

0.5.6-rc7 is a horrible (and horrible old) version of OLSRd. I would
suggest moving to 0.6.2 or 0.6.3 as soon as possible, otherwise you will
have to deal with a LOT of bugs which have been fixed for years.

> I downloaded the most recent src from www.olsr.org, and there seems
> to be some differences in the config file, and it was not able to
> find the plugins properly.

Can you maybe describe what kind of plugins are you using and how your 
configuration file looks (and whats the output of olsrd is when you 
start with your old configuration file)?

> In any case, on looking for others who have had problems with olsr
> and looping, it seems this has been a problem for some time. I don't
> know why I didn't see this before when I was using it more, but
> perhaps most of my 'meshes' were in fact, rather linear or trees and
> did not contain loops.
>
> Does anyone have some suggestions about dealing with mesh

(Continue reading)

headers

John Clark | 26 Jun 2012 20:29

Re: OLSR mesh with loop potential... having problems.

Thanks all,

I updated my olsrd to the 6.3 version that's on the download page. Compiled ok, and did an install and
'install_libs', which seems to have cured my problem with loading the plugins.

I used the default installed olsrd.conf, and only changed the interfaces list.

My net/mesh has 4 nodes, and looks like a 'wheel', where the nodes have direct connection to a 'central
node', for 3 direct connections, and there are 2 other connections that connect the 3 other 'outer' nodes.

When I have all the network connections made, the route tables seem to 'work'...

From the point of view of one of the outher nodes, call it Outer-1, when all is connected, I 'see' the
Central-Node in the route table as would be expected.
And networks on the Central-Node, that are not directly connected to Outer-1, are listed as 2 hops away.

I also see Outer-2, directly connected to Outer-1 in the 'LINKS' section.

When I unplug (simulating nework outage...) the direct connect between Outer-1 and Central-Node,
Outer-1 route tables do not update to forward to Central-Node, via Outer-2.

I'll try to give a text diagram...

Outer-1 <--Link 4-----> Outer-2 <----Link 5---> Outer-3
| | |
Link 1 Link 2 Link 3
| | |
| | |
[ Central-node three separate interfaces connect to the 3 outer node units ]

(Continue reading)

headers

Markus Kittenberger | 26 Jun 2012 21:06

Re: OLSR mesh with loop potential... having problems.

On Tue, Jun 26, 2012 at 8:29 PM, John Clark <jeclark2006 <at> aim.com> wrote:

Thanks all,

I updated my olsrd to the 6.3 version that's on the download page. Compiled ok, and did an install and 'install_libs', which seems to have cured my problem with loading the plugins.

(you could have done an install_all)

I used the default installed olsrd.conf, and only changed the interfaces list.

try to use a new olsrd.conf (i.e an absolutely empty one except the protcol version ipv4 or ipv6 and the interface names)

this will use default settings for everything, and might cure your actual problems,..

As a note, in the past, upon reflection, my 'meshes' have been either linear chains or trees, and there have bee no internal loops.

olsrd works on such topologies, as infact such topologies are what it is meant to operate on.

(it definetely does not need the networks to be a tree,..)

I'd hate to have to put in some sort of 'sensing code' to sense the connectivity on the various routes, and bring interfaces up, or shut them down, to force the mesh to be a tree or chain...

no need for that,..

Markus

--

-- 
Olsr-users mailing list
Olsr-users <at> lists.olsr.org
https://lists.olsr.org/mailman/listinfo/olsr-users

headers

John Clark | 26 Jun 2012 21:27

Re: OLSR mesh with loop potential... having problems.

On Jun 26, 2012, at 12:06 PM, Markus Kittenberger wrote:

On Tue, Jun 26, 2012 at 8:29 PM, John Clark <jeclark2006 <at> aim.com> wrote:
Thanks all,

I updated my olsrd to the 6.3 version that's on the download page. Compiled ok, and did an install and 'install_libs', which seems to have cured my problem with loading the plugins.

(you could have done an install_all)

I used the default installed olsrd.conf, and only changed the interfaces list.

try to use a new olsrd.conf (i.e an absolutely empty one except the protcol version ipv4 or ipv6 and the interface names)

I believe I'm using that conf, all I changed was the interface list.

I'll recheck that...

this will use default settings for everything, and might cure your actual problems,..

As a note, in the past, upon reflection, my 'meshes' have been either linear chains or trees, and there have bee no internal loops.

olsrd works on such topologies, as infact such topologies are what it is meant to operate on.
(it definetely does not need the networks to be a tree,..)

I've been working on this more, and will try to give more details as they come to me, but at the moment, I have the problem when the one link is removed, problems arise.

Thanks,

John Clark.

--

-- 
Olsr-users mailing list
Olsr-users <at> lists.olsr.org
https://lists.olsr.org/mailman/listinfo/olsr-users

headers

Érico Porto | 26 Jun 2012 23:54

Re: OLSR mesh with loop potential... having problems.

I think he meant a file containing the following lines

IpVersion 4

Interface "wlan0"

{

}

Érico V. Porto

On Tue, Jun 26, 2012 at 4:27 PM, John Clark <jeclark2006 <at> aim.com> wrote:

On Jun 26, 2012, at 12:06 PM, Markus Kittenberger wrote:

On Tue, Jun 26, 2012 at 8:29 PM, John Clark <jeclark2006 <at> aim.com> wrote:
Thanks all,

I updated my olsrd to the 6.3 version that's on the download page. Compiled ok, and did an install and 'install_libs', which seems to have cured my problem with loading the plugins.

(you could have done an install_all)

I used the default installed olsrd.conf, and only changed the interfaces list.

try to use a new olsrd.conf (i.e an absolutely empty one except the protcol version ipv4 or ipv6 and the interface names)

I believe I'm using that conf, all I changed was the interface list.

I'll recheck that...

this will use default settings for everything, and might cure your actual problems,..

As a note, in the past, upon reflection, my 'meshes' have been either linear chains or trees, and there have bee no internal loops.

olsrd works on such topologies, as infact such topologies are what it is meant to operate on.
(it definetely does not need the networks to be a tree,..)

I've been working on this more, and will try to give more details as they come to me, but at the moment, I have the problem when the one link is removed, problems arise.

Thanks,

John Clark.

--
Olsr-users mailing list
Olsr-users <at> lists.olsr.org
https://lists.olsr.org/mailman/listinfo/olsr-users

--

-- 
Olsr-users mailing list
Olsr-users <at> lists.olsr.org
https://lists.olsr.org/mailman/listinfo/olsr-users

headers

Vigneswaran R | 5 Oct 2012 07:14

Re: OLSR mesh with loop potential... having problems.

Dear all,

Recently we also noticed loops in the routes established by OLSRv1, on a partial mesh topology (with ~20 machines). We are using olsr 0.6.3 with default configuration file and only Interface line has updated. Not using any plugins. As of now wired connections and at the deployment will be using wireless radios.

Due to the loop, the packets sent through one Interface and received on another. ssh, ping worked fine regardless of that. However, we realized the loop when 'nmap' erroneously reported a node as down.

One point to note here is, when we conducted the test we used 10.0.0.0/8 subnet for all the nodes, so that we can do random reconnects to simulate node movements. However, Henning suggested to put different Interfaces of the same router into different subnets (in his reply on another thread). So, when we do the tests again we will see whether that has any impact on loops.

However, whether loop in routes are expected in general, for such topologies? or by doing some configuration changes can we avoid that?

Thank you.

Regards,
Vignesh

On Wednesday 27 June 2012 03:24 AM, Érico Porto wrote:

I think he meant a file containing the following lines

IpVersion 4

Interface "wlan0"

{

}

Érico V. Porto

On Tue, Jun 26, 2012 at 4:27 PM, John Clark <jeclark2006 <at> aim.com> wrote:

On Jun 26, 2012, at 12:06 PM, Markus Kittenberger wrote:

On Tue, Jun 26, 2012 at 8:29 PM, John Clark <jeclark2006 <at> aim.com> wrote:
Thanks all,

I updated my olsrd to the 6.3 version that's on the download page. Compiled ok, and did an install and 'install_libs', which seems to have cured my problem with loading the plugins.

(you could have done an install_all)

I used the default installed olsrd.conf, and only changed the interfaces list.

try to use a new olsrd.conf (i.e an absolutely empty one except the protcol version ipv4 or ipv6 and the interface names)

I believe I'm using that conf, all I changed was the interface list.

I'll recheck that...

this will use default settings for everything, and might cure your actual problems,..

As a note, in the past, upon reflection, my 'meshes' have been either linear chains or trees, and there have bee no internal loops.

olsrd works on such topologies, as infact such topologies are what it is meant to operate on.
(it definetely does not need the networks to be a tree,..)

I've been working on this more, and will try to give more details as they come to me, but at the moment, I have the problem when the one link is removed, problems arise.

Thanks,

John Clark.

--
Olsr-users mailing list
Olsr-users <at> lists.olsr.org
https://lists.olsr.org/mailman/listinfo/olsr-users

--

-- 
Olsr-users mailing list
Olsr-users <at> lists.olsr.org
https://lists.olsr.org/mailman/listinfo/olsr-users

headers

Markus Kittenberger | 6 Oct 2012 21:23

Re: OLSR mesh with loop potential... having problems.

On Fri, Oct 5, 2012 at 7:14 AM, Vigneswaran R <vignesh <at> atc.tcs.com> wrote:

Dear all,

Recently we also noticed loops in the routes established by OLSRv1, on a partial mesh topology (with ~20 machines). We are using olsr 0.6.3 with default configuration file and only Interface line has updated. Not using any plugins. As of now wired connections and at the deployment will be using wireless radios.

Due to the loop, the packets sent through one Interface and received on another.

you mean return packets just took another route as the packets towards?

this is normal, and you need e.g just 4 nodes with all having 2 etx-1.0 links, aligned in a square, to have a 50% chance to get this, and imho there is nothing wrong about this,..

Markus

ssh, ping worked fine regardless of that. However, we realized the loop when 'nmap' erroneously reported a node as down.

One point to note here is, when we conducted the test we used 10.0.0.0/8 subnet for all the nodes, so that we can do random reconnects to simulate node movements. However, Henning suggested to put different Interfaces of the same router into different subnets (in his reply on another thread). So, when we do the tests again we will see whether that has any impact on loops.

However, whether loop in routes are expected in general, for such topologies? or by doing some configuration changes can we avoid that?

Thank you.

Regards,
Vignesh

On Wednesday 27 June 2012 03:24 AM, Érico Porto wrote:

I think he meant a file containing the following lines

IpVersion 4

Interface "wlan0"

{

}

Érico V. Porto

On Tue, Jun 26, 2012 at 4:27 PM, John Clark <jeclark2006 <at> aim.com> wrote:

On Jun 26, 2012, at 12:06 PM, Markus Kittenberger wrote:

On Tue, Jun 26, 2012 at 8:29 PM, John Clark <jeclark2006 <at> aim.com> wrote:
Thanks all,

I updated my olsrd to the 6.3 version that's on the download page. Compiled ok, and did an install and 'install_libs', which seems to have cured my problem with loading the plugins.

(you could have done an install_all)

I used the default installed olsrd.conf, and only changed the interfaces list.

try to use a new olsrd.conf (i.e an absolutely empty one except the protcol version ipv4 or ipv6 and the interface names)

I believe I'm using that conf, all I changed was the interface list.

I'll recheck that...

this will use default settings for everything, and might cure your actual problems,..

As a note, in the past, upon reflection, my 'meshes' have been either linear chains or trees, and there have bee no internal loops.

olsrd works on such topologies, as infact such topologies are what it is meant to operate on.
(it definetely does not need the networks to be a tree,..)

I've been working on this more, and will try to give more details as they come to me, but at the moment, I have the problem when the one link is removed, problems arise.

Thanks,

John Clark.

--
Olsr-users mailing list
Olsr-users <at> lists.olsr.org
https://lists.olsr.org/mailman/listinfo/olsr-users

--
Olsr-users mailing list
Olsr-users <at> lists.olsr.org
https://lists.olsr.org/mailman/listinfo/olsr-users

--

-- 
Olsr-users mailing list
Olsr-users <at> lists.olsr.org
https://lists.olsr.org/mailman/listinfo/olsr-users

headers

Vigneswaran R | 8 Oct 2012 07:10

Re: OLSR mesh with loop potential... having problems.

On Sunday 07 October 2012 12:53 AM, Markus Kittenberger wrote:

On Fri, Oct 5, 2012 at 7:14 AM, Vigneswaran R <vignesh <at> atc.tcs.com> wrote:

Dear all,

Recently we also noticed loops in the routes established by OLSRv1, on a partial mesh topology (with ~20 machines). We are using olsr 0.6.3 with default configuration file and only Interface line has updated. Not using any plugins. As of now wired connections and at the deployment will be using wireless radios.

Due to the loop, the packets sent through one Interface and received on another.

you mean return packets just took another route as the packets towards?

Yes, exactly.

this is normal, and you need e.g just 4 nodes with all having 2 etx-1.0 links, aligned in a square, to have a 50% chance to get this, and imho there is nothing wrong about this,..

Ok. Thank you for the clarification.

Regards,
Vignesh

Markus

ssh, ping worked fine regardless of that. However, we realized the loop when 'nmap' erroneously reported a node as down.

One point to note here is, when we conducted the test we used 10.0.0.0/8 subnet for all the nodes, so that we can do random reconnects to simulate node movements. However, Henning suggested to put different Interfaces of the same router into different subnets (in his reply on another thread). So, when we do the tests again we will see whether that has any impact on loops.

However, whether loop in routes are expected in general, for such topologies? or by doing some configuration changes can we avoid that?

Thank you.

Regards,
Vignesh

On Wednesday 27 June 2012 03:24 AM, Érico Porto wrote:

I think he meant a file containing the following lines

IpVersion 4

Interface "wlan0"

{

}

Érico V. Porto

On Tue, Jun 26, 2012 at 4:27 PM, John Clark <jeclark2006 <at> aim.com> wrote:

On Jun 26, 2012, at 12:06 PM, Markus Kittenberger wrote:

On Tue, Jun 26, 2012 at 8:29 PM, John Clark <jeclark2006 <at> aim.com> wrote:
Thanks all,

I updated my olsrd to the 6.3 version that's on the download page. Compiled ok, and did an install and 'install_libs', which seems to have cured my problem with loading the plugins.

(you could have done an install_all)

I used the default installed olsrd.conf, and only changed the interfaces list.

try to use a new olsrd.conf (i.e an absolutely empty one except the protcol version ipv4 or ipv6 and the interface names)

I believe I'm using that conf, all I changed was the interface list.

I'll recheck that...

this will use default settings for everything, and might cure your actual problems,..

As a note, in the past, upon reflection, my 'meshes' have been either linear chains or trees, and there have bee no internal loops.

olsrd works on such topologies, as infact such topologies are what it is meant to operate on.
(it definetely does not need the networks to be a tree,..)

I've been working on this more, and will try to give more details as they come to me, but at the moment, I have the problem when the one link is removed, problems arise.

Thanks,

John Clark.

--
Olsr-users mailing list
Olsr-users <at> lists.olsr.org
https://lists.olsr.org/mailman/listinfo/olsr-users

--
Olsr-users mailing list
Olsr-users <at> lists.olsr.org
https://lists.olsr.org/mailman/listinfo/olsr-users

--

-- 
Olsr-users mailing list
Olsr-users <at> lists.olsr.org
https://lists.olsr.org/mailman/listinfo/olsr-users