headers
Mehmet Akcin | 1 Aug 2009 08:30

[quagga-users 10986] Ospf6 and ECMP Behaviour

Hello,

I am currently running Quagga 0.99.14 on CentOS 5.3 and experiencing a
weird problem. I have 8 servers ecmp load balancing the total load,
and everything is working fine with ipv4 (in terms of load balancing
etc...) on ipv6 everything seems to be working fine but no ecmp load
balancing, simply just one route gets the all traffic, no interface
errors though.

Here is ospf6.conf

!
password z3bra
log file /var/log/quagga/ospf6d.log
!
debug ospf6 lsa unknown
!
interface eth0
 ipv6 ospf6 cost 1
 ipv6 ospf6 priority 0
!
interface lo
 ipv6 ospf6 cost 1
 ipv6 ospf6 priority 0
!
router ospf6
 router-id 69.44.23.11
 interface eth0 area 0.0.0.0
 interface lo area 0.0.0.0
!
(Continue reading)

Permalink | Reply | 0 comments |
headers
Adrian Ban | 1 Aug 2009 10:55
Picon

[quagga-users 10987] Re: bgp multiple instance

Ok. I understand now why that project was negated. I look a bit in those sources and first think was: damn .. another patch for kernel? And after that it cames me that ideea with multi-routing tables that i thing it is almost the same thing.

I'm not an advanced programmer, but i can help you if i have the documentation or some starting points for developing.

Of course. It will a pleasure to help with anything :). Offtopic for general information i'm maintaining the realm project for Quagga for last versions :).

James Leu wrote:
I'm the developer behind the Linux VRF project. Since the integration of network namespaces in the Linux kernel, the core of the linux VRF project has been negated. As soon as I finish cleaning up my quagga nexthop and netlink development so Paul can entertain integrating it, I plan to redefined the linux-vrf project to focus more on the userland aspect of VRFs and less on the kernel implementation. Let me know if you would like to participate in that work. On Fri, Jul 31, 2009 at 04:25:36PM +0300, Adrian Ban wrote:
Yes, i looked at this project, but first is outdated for a long timp: kernel 2.6.8. Second i don't find any implementation of VRF for Quagga. It is a project named MPLS for Linux that implements, but there is dificult to apply. Everton Marques wrote:
On Thu, Jul 30, 2009 at 10:11 PM, Adrian Ban wrote:
This is an intresting question. I'm intrested too about of implementation of VRF on linux via Quagga. If any body tell us if there is any project to implement a VRF function in Quagga, i'll apreciate that.
Have you looked at Linux VRF ? http://linux-vrf.sourceforge.net/ Everton
_______________________________________________ Quagga-users mailing list Quagga-users-UOy77sIEA+cAd7ICUelF/Q@public.gmane.org http://lists.quagga.net/mailman/listinfo/quagga-users 

_______________________________________________
Quagga-users mailing list
Quagga-users@...
http://lists.quagga.net/mailman/listinfo/quagga-users
Permalink | Reply | 0 comments |
headers
Tore Anderson | 4 Aug 2009 08:58
Picon

[quagga-users 10988] Interface configuration handling in ospf6

Good morning all,

I'm building a "router on a stick" using Quagga/OSPF with keepalived for
VRRP.  In IPv4, I can do the following:

router ospf
  passive-interface default
  no passive-interface eth0
  network 192.168.0.0/16 area 0.0.0.0

...which works very well as I do not have to make any changes to the
Quagga configuration if I add a new VLAN interface with a VIP in the
192.168.0.0/16 network - it gets picked up automatically and announced
to the upstream router(s) attached to eth0.  Perfect.

With ospf6, however, I'm left with this:

interface vlan1
  ipv6 ospf6 passive
!
[...]
!
interface vlanNNN
  ipv6 ospf6 passive
!
router ospf6
  area 0.0.0.0 range 2001:db8::/32
  interface eth0 area 0.0.0.0
  interface vlan1 area 0.0.0.0
  [...]
  interface vlanNNN area 0.0.0.0

So I need to set both "interface foo area 0.0.0.0" under "router ospf6"
and "ipv6 ospf6 passive" for every single VLAN interface in order to
properly activate OSPFv3 operation on the access VLANs.  Is there
another way to configure this (that I've missed) so that it's all
automatic, like it is for OSPFv2?

On a related note, when I enter the "interface foo area 0.0.0.0" under
"router ospf6", the following default settings are added under
"interface foo"

 ipv6 ospf6 cost 1
 ipv6 ospf6 hello-interval 10
 ipv6 ospf6 dead-interval 40
 ipv6 ospf6 retransmit-interval 5
 ipv6 ospf6 priority 1
 ipv6 ospf6 transmit-delay 1
 ipv6 ospf6 instance-id 0

While not a big problem, these lines bloat the configuration file a lot
(especially since they're repeated for every OSPFv3-enabled interface).
 Is there a way to make them not appear as long as they're left at their
defaults?  This seems to be how Quagga behaves in most other places, and
it keeps the configuration file neat and readable.

Best regards,
--

-- 
Tore Anderson
Redpill Linpro AS - http://www.redpill-linpro.com/
Tel: +47 21 54 41 27
Permalink | Reply | 0 comments |
headers
Ricardo Oliveira | 5 Aug 2009 03:31
Picon

[quagga-users 10989] Cluster-list and originator-id in MRT RIB dumps

Hi,

Just wanted to check if the current MRT RIB dumps from quagga include  
the cluster-list and originator-id BGP attributes? I'm dealing w/ some  
data from a BGP session where the updates contain these attributes,  
but they are omitted from the RIBs... maybe i'm missing some config  
lines at the quagga side..

Thanks,

--Ricardo
Permalink | Reply | 0 comments |
headers
Maciej Drobniuch | 5 Aug 2009 09:49
Picon

[quagga-users 10990] [OSPFv2] Problem with neighbor adjacency

Hi all!
I have five ethernet interfaces(eth0-eth4). eth0 is an integrated card(via) and eth1-eth4 is an all in one gigabit ethernet card(4 ports, probably based on rt8139). OSPF runs on eth0 and eth4. The problem is that quagga can't see neigboards on eth4, the adjacency works perfectly on eth0.
I've changed in the /etc/udev/rules/75-network... the interfaces attached to eth0 and eth4, i've switched them physicly and logicaly, so the configuration files are the same. Now the adjacency is working on eth4 and not on eth0. What is interesting the neighboar of the interface that isn't working sees the quagga router in INIT state.

My configuration:
zebra.conf
! -*- zebra -*-
!
! zebra sample configuration file
!
! $Id: zebra.conf.sample,v 1.1.1.1 2002/12/13 20:15:30 paul Exp $
!
hostname localhost
password sdfsdf
!
! Interface's description.
!
!interface lo
! description test of desc.
!
!interface sit0
! multicast

interface eth0
 ip address 82.0.0.245/30
 multicast

interface eth4
 ip address 82.0.0.210/30
 multicast

!interface eth2
! ip address 192.168.0.1/24
! multicast

!
! Static default route sample.
!
#ip route 0.0.0.0/0 83.13.157.161
!

log file zebra.log


ospfd.conf
! -*- ospf -*-
!
! OSPFd sample configuration file
!
!
hostname ospfd
password tereferekuku
!enable password please-set-at-here
!
router ospf
  ospf router-id 82.0.0.210
  network 82.0.0.244/30 area 0
  network 82.0.0.210/30 area 0
  !default-information originate always
  !redistribute static
  redistribute connected
!
!interface eth4
! ip ospf cost 10
! ip ospf hello-interval 10
! ip ospf dead-interval 40
! ip ospf retransmit-interval 5
! ip ospf priority 1
! ip ospf transmit-delay 1
!
log stdout
log file ospfd.log

ospfd.log:
2009/08/05 04:48:56 OSPF: OSPFd 0.99.13 starting: vty <at> 2604
2009/08/05 04:48:56 OSPF: interface 82.0.0.245 [5] join AllSPFRouters Multicast group.
2009/08/05 04:48:56 OSPF: interface 82.0.0.210 [6] join AllSPFRouters Multicast group.
2009/08/05 04:48:59 OSPF: ospfTrapNbrStateChange trap sent: 82.160.22.209 now Init/DROther
2009/08/05 04:49:02 OSPF: DR-Election[1st]: Backup 82.0.0.210
2009/08/05 04:49:02 OSPF: DR-Election[1st]: DR     82.0.0.209
2009/08/05 04:49:02 OSPF: DR-Election[2nd]: Backup 82.0.0.210
2009/08/05 04:49:02 OSPF: DR-Election[2nd]: DR     82.0.0.209
2009/08/05 04:49:02 OSPF: ospfTrapIfStateChange trap sent: 82.0.0.210 now Backup
2009/08/05 04:49:02 OSPF: interface 82.0.0.210 [6] join AllDRouters Multicast group.
2009/08/05 04:49:02 OSPF: DR-Election[1st]: Backup 82.00.0.210
2009/08/05 04:49:02 OSPF: DR-Election[1st]: DR     82.0.0.209
2009/08/05 04:49:02 OSPF: Packet[DD]: Neighbor 192.168.127.1 Negotiation done (Slave).
2009/08/05 04:49:07 OSPF: nsm_change_state(192.168.127.1, Loading -> Full): scheduling new router-LSA origination
2009/08/05 04:49:36 OSPF: DR-Election[1st]: Backup 82.0.0.245
2009/08/05 04:49:36 OSPF: DR-Election[1st]: DR     82.0.0.245
2009/08/05 04:49:36 OSPF: DR-Election[2nd]: Backup 0.0.0.0
2009/08/05 04:49:36 OSPF: DR-Election[2nd]: DR     82.0.0.245
2009/08/05 04:49:36 OSPF: ospfTrapIfStateChange trap sent: 82.0.0.245 now DR
2009/08/05 04:49:36 OSPF: interface 82.0.0.245 [5] join AllDRouters Multicast group.

192.168.127.1 is the router behind eth4(via card)
82.0.0.246 is the router behind(eth0) the combo card based on rtl8139

Could someone tell me howto debug logs and put out the useful stuf?
Thanks for suggestions and sorry for my lame eng.
Best wishes
Maciej Drobniuch

_______________________________________________
Quagga-users mailing list
Quagga-users@...
http://lists.quagga.net/mailman/listinfo/quagga-users
Permalink | Reply | 0 comments |
headers
For@ll | 5 Aug 2009 15:16

[quagga-users 10991] quagga and 2 isp - how set operator A to a primary

Hi,

I have two independent operators, lets name then A (10Mgbit / s) and B 
(2Mgbit / s). I'd like to redirect all traffic coming in and outgoing 
the operator A, while B only as a backup.
Last I noticed that customers who have a data link from operator B,if 
they are send me something, it's used a operator B not A.
How can I force it in the settings to make sure everything was done 
Quagga on operator A.

Regards,

For <at> ll
Permalink | Reply | 2 comments |
headers
Jack Carrozzo | 5 Aug 2009 15:45

[quagga-users 10992] Re: quagga and 2 isp - how set operator A to a primary

What routing protocols are you using... do you speak BGP to both providers?

Since you didn't mention it, I'll assume no... One way to accomplish
this (and make it a little more fault tolerant) is to add a gateway
for each ISP. Drop 0/0 into OSPF from each box, but set the metrics
such that ISP A is chosen over B. This way A will be chosen by your
internal gateways, but will move over to B in the event A fails. You
can set your hello-interval quite low and get sub-second convergence.

-Jack Carrozzo

On Wed, Aug 5, 2009 at 9:16 AM, For <at> ll<forall@...> wrote:
> Hi,
>
> I have two independent operators, lets name then A (10Mgbit / s) and B
> (2Mgbit / s). I'd like to redirect all traffic coming in and outgoing
> the operator A, while B only as a backup.
> Last I noticed that customers who have a data link from operator B,if
> they are send me something, it's used a operator B not A.
> How can I force it in the settings to make sure everything was done
> Quagga on operator A.
>
>
> Regards,
>
> For <at> ll
>
> _______________________________________________
> Quagga-users mailing list
> Quagga-users@...
> http://lists.quagga.net/mailman/listinfo/quagga-users
>
Permalink | Reply | 0 comments |
headers
Jack Carrozzo | 5 Aug 2009 16:03

[quagga-users 10993] Re: quagga and 2 isp - how set operator A to a primary

Oh, then just set the weights. See below if you need help:

http://www.cisco.com/en/US/docs/internetworking/technology/handbook/bgp.html

-Jack Carrozzo

On Wed, Aug 5, 2009 at 9:59 AM, For <at> ll<forall@...> wrote:
> Jack Carrozzo pisze:
>>
>> What routing protocols are you using... do you speak BGP to both
>> providers?
>>
>> Since you didn't mention it, I'll assume no... One way to accomplish
>> this (and make it a little more fault tolerant) is to add a gateway
>> for each ISP. Drop 0/0 into OSPF from each box, but set the metrics
>> such that ISP A is chosen over B. This way A will be chosen by your
>> internal gateways, but will move over to B in the event A fails. You
>> can set your hello-interval quite low and get sub-second convergence.
>>
>>
>
> Hi,
>
> I using a bgp procotol with both providers, and I was looking how to resolve
> my problem.
> All trafic input and output with my first providers which name (A).
>
> For <at> ll
>
Permalink | Reply | 0 comments |
headers
Adrian Ban | 5 Aug 2009 16:24
Picon

[quagga-users 10994] Re: quagga and 2 isp - how set operator A to a primary

You are using BGP for interconect with those operators? If the answer is 
yes, look for  set  as-path prepend in route-map for outgoing routes. 
Put more prepands on one of operators you'll balance the traffic for 
incomming traffic.

Also if you want to prefer the routes from A instead B, check for set 
local-preference for incoming routes.

BTW: don't confuse incoming routes with incoming traffic or outgoing 
routes with outgoing traffic!! the incoming routes are routes installed 
in your box for outgoing traffic! The outgoing routes are routes to your 
networks installed on operators A and B for giving the internet traffic 
to you, aka your incoming traffic.

Best regards,
Adrian

For <at> ll wrote:
> Hi,
>
> I have two independent operators, lets name then A (10Mgbit / s) and B 
> (2Mgbit / s). I'd like to redirect all traffic coming in and outgoing 
> the operator A, while B only as a backup.
> Last I noticed that customers who have a data link from operator B,if 
> they are send me something, it's used a operator B not A.
> How can I force it in the settings to make sure everything was done 
> Quagga on operator A.
>
>
> Regards,
>
> For <at> ll
>
> _______________________________________________
> Quagga-users mailing list
> Quagga-users@...
> http://lists.quagga.net/mailman/listinfo/quagga-users
>
Permalink | Reply | 0 comments |
headers
Anirban Sinha | 6 Aug 2009 00:30
Favicon

[quagga-users 10995] Linux kernel support for large FIB

Hi All:

We need to support about 512K routes in the kernel FIB for the BGP
routing software built in-house. We are using off-the-shelf linux kernel
2.6.26 from linux-mips.org for our mips platform. Since you guys are
already working on this BGP routing space, I am wondering if you guys
can answer two of my questions:

(a) Is the Linux kernel able to support the large FIB size? If not, are
there any patches available for this support?
(b) Are there any known performance implications for having a  very
large number of routing entries in the kernel? We are uing about 4 gigs
of memory (RAM).

I will be really grateful if any of you guys can throw some insights
into these questions. 

Thanks for the time.

Cheers,

Ani
Permalink | Reply | 6 comments |

Gmane