clownix at clownix
Sep 18, 2011, 3:03 PM
Post #3 of 3
(348 views)
Permalink
|
Kernel version for lxc is the host's kernel version, 2.6.39.3 in my host,
and lxc is lxc-0.7.5.tar.gz (it is compiled with the rest of cloonix).
The problem is probably due to the ip stack isolation not complete between
the host and all the lxc guests, that means that when you do a traceroute
in your virtual network, the host's kernel sees many icmp replies and the
icmp_ratemask acts. Putting 0 in /proc/sys/net/ipv4/icmp_ratemask in all
the guests takes this problem away. But there seem to be another parameter
I have not found that acts on the number of multicast that can be
received.
If you want to follow this problem, it is better to post at the lxc list.
> Hi Vince,
>
> Any shot there is some kind of physical discontinuity (or multicast
> problem) in the lxc container or host kernel? Can you please include
> the kernel / lxc versions and I'll try to spin them up? (Does unicast
> stay working thruout? )
>
> --Adrian
>
> On Sun, Sep 18, 2011 at 9:46 AM, <clownix [at] clownix> wrote:
>> Hello,
>> I have 2 exactly identical topologies running ospf: one based on uml
>> virtual machines and one based on lxc virtual machines.
>>
>> The 2 topologies are available at clownix.net.
>>
>> My problem is that the lxc based demo does not work well, it stays
>> stable
>> only for moments before reloading LS.
>>
>> I have a message that does not exist in the uml demo, it is probably the
>> cause of loss of hellos:
>>
>> V3 Membership Report / Leave group 224.0.0.6 / Leave group 224
>>
>> What could cause ospf to leave the multicast group now and again?
>>
>> Note that I had previously another problem only for lxc:
>> to have a command traceroute that worked, I had to do:
>> echo 0 > /proc/sys/net/ipv4/icmp_ratemask in all my lxc machines.
>>
>> This is because in lxc the rate counter sum all values of all virtual
>> machine lxc, only one host kernel being used for all machines.
>>
>> Is there some sort of rate counter for multicast messages?
>> Thanks for your help.
>> Vincent
>>
>>
>> _______________________________________________
>> Quagga-users mailing list
>> Quagga-users [at] lists
>> http://lists.quagga.net/mailman/listinfo/quagga-users
>>
>
_______________________________________________
Quagga-users mailing list
Quagga-users [at] lists
http://lists.quagga.net/mailman/listinfo/quagga-users
|
