T
Phil K
Is it considered to be simplier than the architecture I talked about before ?
more explained at least
Size: a a a
2020 June 11
J
я все сижу на linuxbridge и как-то норм. В чем смысл пересаживания на OVN?
Distributed SNAT, исчезает лапша из линукс бриджей, ovs бриджей и чудо-агентов, которые получают сообщения через rabbitmq и потом через Openflow потоки создают.
Там типа всё сразу на openflow.
Там типа всё сразу на openflow.
T
Phil K
But how it will handle when the network node is dead ?
you should deploy an openstack environment and try it for yourself
no other way to really understand how it works
no other way to really understand how it works
J
Phil K
Is it considered to be simplier than the architecture I talked about before ?
It's a deployment guide for architecture I have mentioned:
https://docs.openstack.org/liberty/networking-guide/scenario-dvr-ovs.html
https://docs.openstack.org/liberty/networking-guide/scenario-dvr-ovs.html
AM
Distributed SNAT, исчезает лапша из линукс бриджей, ovs бриджей и чудо-агентов, которые получают сообщения через rabbitmq и потом через Openflow потоки создают.
Там типа всё сразу на openflow.
Там типа всё сразу на openflow.
а это вообще как-то тшутить можно или смотреть как на magic blackbox?
PK
you should deploy an openstack environment and try it for yourself
no other way to really understand how it works
no other way to really understand how it works
I agree with you
AM
мне тут cilium с ebpf хватает 😁
J
а это вообще как-то тшутить можно или смотреть как на magic blackbox?
flow таблицы смотреть и базы на контроллерах.
Вполне можно, я думаю, хотя и не так привычно.
Вполне можно, я думаю, хотя и не так привычно.
AM
flow таблицы смотреть и базы на контроллерах.
Вполне можно, я думаю, хотя и не так привычно.
Вполне можно, я думаю, хотя и не так привычно.
ну ок, может как-то пересядем, хотя как я понимаю это для больших окружений, а у нас пока почти все во ВЛАН сидят
T
я все сижу на linuxbridge и как-то норм. В чем смысл пересаживания на OVN?
в каком режиме у тебя роутинг реализован?
T
а, если ВЛАН то понятно
AM
в каком режиме у тебя роутинг реализован?
3 - l3-agent-ha, на 5 роутеров хватает
PK
Спасибо всем !
T
Phil K
Спасибо всем !
ni puha
J
Phil K
But how it will handle when the network node is dead ?
Basically, idea is the same for most popular Neutron network scenarios that support DVR.
You have an instance of virtual router on multiple nodes (in OVN they are called chassis).
At the same time only one instance on a single node is used to serve N/S traffic for SNAT VMs. Which instance of the virtual router will be picked is determined by their priority. When one node or one router instance goes down instance of the same router with second highest priority becomes active. And so on.
You have an instance of virtual router on multiple nodes (in OVN they are called chassis).
At the same time only one instance on a single node is used to serve N/S traffic for SNAT VMs. Which instance of the virtual router will be picked is determined by their priority. When one node or one router instance goes down instance of the same router with second highest priority becomes active. And so on.
J
Basically, idea is the same for most popular Neutron network scenarios that support DVR.
You have an instance of virtual router on multiple nodes (in OVN they are called chassis).
At the same time only one instance on a single node is used to serve N/S traffic for SNAT VMs. Which instance of the virtual router will be picked is determined by their priority. When one node or one router instance goes down instance of the same router with second highest priority becomes active. And so on.
You have an instance of virtual router on multiple nodes (in OVN they are called chassis).
At the same time only one instance on a single node is used to serve N/S traffic for SNAT VMs. Which instance of the virtual router will be picked is determined by their priority. When one node or one router instance goes down instance of the same router with second highest priority becomes active. And so on.
Implementations are different, but the idea is the same.
PK
Basically, idea is the same for most popular Neutron network scenarios that support DVR.
You have an instance of virtual router on multiple nodes (in OVN they are called chassis).
At the same time only one instance on a single node is used to serve N/S traffic for SNAT VMs. Which instance of the virtual router will be picked is determined by their priority. When one node or one router instance goes down instance of the same router with second highest priority becomes active. And so on.
You have an instance of virtual router on multiple nodes (in OVN they are called chassis).
At the same time only one instance on a single node is used to serve N/S traffic for SNAT VMs. Which instance of the virtual router will be picked is determined by their priority. When one node or one router instance goes down instance of the same router with second highest priority becomes active. And so on.
Thank you for this message it starts to clarify my mind
PK
When I'll need to deploy I must specify the weights ?
J
Phil K
When I'll need to deploy I must specify the weights ?
I don't know, unfortunately. Possibly, both manual and automatic priority assignment should be supported.