[Openstack-operators] What is the best production kernel, openvswitch combination?

Joseph Breu breu at breu.org
Tue Feb 11 16:27:40 UTC 2014


Hi Jesse,

We have had success with 3.2.0-58 and OVS 1.10.2.  We too have experienced kernel panics with stock OVS and 3.2.0-58

---
Joseph Breu
Deployment Engineer
Rackspace Private Cloud
210-312-3508

On Feb 11, 2014, at 10:20 AM, Simon Pasquier <simon.pasquier at bull.net> wrote:

> Hi Jesse,
> 
> I've run into the same kernel panic you experienced [1] and openvswitch-1.4.6 didn't help either. I don't have the issue with kernel 3.2.0-56 so I'm sticking with that version.
> I've no time to investigate further in the short term but I'll keep you posted.
> 
> HTH,
> 
> Simon
> 
> [1] https://bugs.launchpad.net/openstack-cisco/+bug/1260334
> 
> On 09/02/2014 12:08, Jesse Pretorius wrote:
>> Hi everyone,
>> 
>> We have a production cloud system running on Ubuntu 12.04 LTS with
>> Openstack Grizzly from the Ubuntu Cloud Archive. We're using the stock
>> Openvswitch from the Precise repo.
>> 
>> Current levels:
>> kernel 3.2.0-54
>> openvswitch 1.4.0-1ubuntu1.5
>> openstack 1:2013.1.4-0ubuntu1~cloud0
>> 
>> Recently I tried doing a dist-upgrade to kernel 3.2.0-58. After the
>> reboot everything seemed fine, then the quantum ovs agent started,
>> initiated the gre tunnels and the kernel panicked. Not good. As it turns
>> out, this has been picked up by others:
>> https://bugs.launchpad.net/openstack-cisco/+bug/1260334
>> https://bugs.launchpad.net/ubuntu/precise/+source/openvswitch/+bug/1262692
>> 
>> So, the simplest solution appeared to be to try updating to
>> openvswitch-1.4.6 from precise-proposed. I tried that (admittedly
>> keeping the kernel version static) on a network node and it borked again.
>> 
>> Alright, so perhaps I should just let sleeping dogs lie and leave the
>> software levels as they are. That'd be great except that as they are I'm
>> experiencing issues with port tagging not working properly for router
>> internal ports and dhcp service ports:
>> https://bugs.launchpad.net/neutron/+bug/1269795
>> 
>> So, I'd like to know from anyone still reading - what levels of kernel,
>> openvswitch and openvswitch-datapath-dkms do you have working perfectly
>> in production?
>> 
>> I'm aware of the potential performance issues with a sub-v1.11 ovs -
>> slow performance is a lesser issue for me right now, and we have other
>> protections against DOS attacks, I'm more concerned with having a stable
>> system that works as it's supposed to.
>> 
>> Ideally I'd like a combination of packages directly from Ubuntu Archive
>> and Cloud Archive - preferably no compilations from source or PPA's. I'm
>> happy to use some packages from Havana Cloud Archive if they come
>> recommended.
>> 
>> So far it looks to me like this may work:
>> kernel 3.11.0-12 - using linux-lts-saucy from Ubuntu Archive
>> openvswitch 1.10.2 - from Havana Cloud Archive
>> openvswitch-datapath-dkms 1.10.2 - from Havana Cloud Archive
>> 
>> Any thoughts, comments or anecdotes would be highly appreciated.
>> 
>> 
>> _______________________________________________
>> OpenStack-operators mailing list
>> OpenStack-operators at lists.openstack.org
>> http://lists.openstack.org/cgi-bin/mailman/listinfo/openstack-operators
>> 
> 
> 
> _______________________________________________
> OpenStack-operators mailing list
> OpenStack-operators at lists.openstack.org
> http://lists.openstack.org/cgi-bin/mailman/listinfo/openstack-operators

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.openstack.org/pipermail/openstack-operators/attachments/20140211/4d5553dd/attachment.html>


More information about the OpenStack-operators mailing list