[Users] Multicast issue with openvz VMs and CTs

Paulo Coghi - Coghi IT paulocoghi at gmail.com
Wed Jan 11 11:33:29 MSK 2023


Jake, does an L4 load balancer solve the problem?

If yes, maybe Katran could help: https://github.com/facebookincubator/katran

On Wed, Jan 11, 2023 at 9:28 AM Paulo Coghi - Coghi IT <paulocoghi at gmail.com>
wrote:

> Thanks Jake for all the valuable information.
>
> I'm following every email.
>
> On Tue, Jan 10, 2023 at 5:05 PM jjs - mainphrame <jjs at mainphrame.com>
> wrote:
>
>> To clarify, the openvz guests can receive multicast traffic from the lan,
>> but they are unable to send multicast traffic to the lan. The multicast
>> packets are dropped on the way out, somewhere between the guest adapter and
>> the host bridge.
>>
>> I'm not seeing any differences in sysctl settings between the ovz hosts
>> and the working hosts, so firewall rules are the likely culprit.
>>
>> I'll continue to chip away at this as time allows and update with any
>> findings.
>>
>> Jake
>>
>>
>> On Sun, Jan 8, 2023 at 5:08 PM jjs - mainphrame <jjs at mainphrame.com>
>> wrote:
>>
>>> I've been doing some testing with ucarp, in debian VMs and containers.
>>>
>>> (ucarp is an implementation of VRRP, a means of providing a highly
>>> available floating virtual IP within a cluster of machines)
>>>
>>> It works fine on proxmox VMs and CTs, and I was hoping to get it working
>>> on openvz, but so far my attempts to get it fully up and running have
>>> failed.
>>>
>>> Basically, both nodes become master, because neither node is seeing the
>>> multicast traffic from the other.
>>>
>>> What is the secret of allowing multicast traffic to pass from openvz VMs
>>> and CTs, onto the lan?
>>>
>>> Jake
>>>
>> _______________________________________________
>> Users mailing list
>> Users at openvz.org
>> https://lists.openvz.org/mailman/listinfo/users
>>
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.openvz.org/pipermail/users/attachments/20230111/5d72174e/attachment-0001.html>


More information about the Users mailing list