[rdo-list] Shipping OVS 2.6?

[John Trowbridge]

On 01/18/2017 07:44 PM, Alan Pevec wrote:
> On Mon, Jan 9, 2017 at 12:26 PM, Assaf Muller <assaf at redhat.com> wrote:
>> On Fri, Jan 6, 2017 at 9:11 AM, Michael Orazi <morazi at redhat.com> wrote:
>>> Additionally I'd be curious to know if we have CI jobs to cover
>>> in-place upgrades so we understand the implications for systems that
>>> are already set up and would like to move to a newer OVS version.
>>
>> Indeed the 2.4 to 2.5 upgrade was tricky.
> 
> And so will be 2.6 :(
> 
> Test upgrade job with pending repo enabled failed:
> http://logs.openstack.org/44/422244/2/experimental/gate-tripleo-ci-centos-7-multinode-upgrades-nv/5c5d59d/console.html
> ...
> ovs-vsctl: unix:/var/run/openvswitch/db.sock: database connection
> failed (No such file or directory)
> 

Technically, this is not only happening on the upgrades job[1]. However,
it is happening in the infra setup stage for the multinode jobs, and the
image from infra has older OVS and we upgrade before running the
ovs-vsctl command.

This was actually pretty easy for me to reproduce by just upgrading OVS
to 2.6 and running `ovs-vsctl --add-br foo`. I get the same error.

[1]
http://logs.openstack.org/44/422244/2/check/gate-tripleo-ci-centos-7-nonha-multinode/1613617/

> Assaf, can you get those who worked on ovs 2.4 -> 2.5 upgrade to look
> at this, if it might sound familiar to them?
> Potential patches in tripleo can be tested against OVS 2.6 in pending
> repo by adding
> Depends-On: I19293cfef76384598511c9b278cdf887eaa5dc5c
> 
> Cheers,
> Alan
>