Ask Your Question

Why is Liberty Neutron DVR so slow?

asked 2016-04-25 12:07:59 -0500

mattd gravatar image

updated 2016-04-25 12:23:43 -0500

sgordon gravatar image

I am seeing login times greater than 1 minute for initial login for instances. I am just wondering if DVR is production ready in Liberty?

I am using a HA Liberty multi node setup with Neutron DVR using OpenVSwitch on the back-end

I verified the setup at each namespace level to ensure that the packets sent via East-West were traversing the network correctly. As far as I can tell it seems to be set up correctly. But seems to be very slow.

edit retag flag offensive close merge delete

2 answers

Sort by ยป oldest newest most voted

answered 2016-04-25 12:24:21 -0500

dbaxps gravatar image

updated 2016-04-25 13:34:44 -0500

Using field as comment
Quoting :-

 I am using a HA Liberty multi node setup with Neutron DVR using OpenVSwitch on the back-end

Please, elaborate how you configured HA on Liberty ?
The best option ( if possible ) your detailed landscape setup to be provided in question.

You are not answering the question HOW you implement HA setting up 3 Controller nodes 
and having single  Network node ?

Mentioned post of Eduardo Gonzalez
doesn't provide any HA solutions , it is just about DVR.

    Generally in HA configuration 3 Controller/Network nodes are supposed to be involved, due to Neutron Server and all neutron agents running on this nodes should have access to MariaGalera DB instances running MultiMaster Synchronous replication  and your answer seems to be completely missing this point. The Galera DB might experience deadlocks due to attempts to update same record on different nodes, which is resolved by selecting a victim transaction and rolling it back.
edit flag offensive delete link more

answered 2016-04-25 12:31:52 -0500

mattd gravatar image

Thank you for your response. I am at the conference at the Ask session, and they wanted me to quickly post a question, so I apologize for being concise. How did you happen to see this message? We are talking about how to improve ask, and I am curious how my question got to you.

I am using a 3 node controller cluster with RabbitMQ cluster, Galera MySQL cluster with a HAProxy sitting out front where all requests are sent to. I am using a single Neutron Network Node, which handles dhcp, then I have all the Compute nodes running Neutron for everything but the dhcp role that is strictly within the neutron node. I am using OpenvSwitch with Vlan tagged interfaces for the br-* bridges. I am trying to use VXLAN on the back-end for the tunnels.

Basically, I have verified the entire chain of N-S traffic from an instance to the internet, but being that it was so slow, I did not spend much time verifying EW traffic.

I used the instructions found here:

Thank you in advance!!

edit flag offensive delete link more

Get to know Ask OpenStack

Resources for moderators

Question Tools

1 follower


Asked: 2016-04-25 12:07:59 -0500

Seen: 320 times

Last updated: Apr 25 '16