[openstack-dev] [Neutron] [horizon] Netaddr 0.7.16 and gate breakage

Ihar Hrachyshka ihrachys at redhat.com
Mon Aug 31 14:57:50 UTC 2015


> On 31 Aug 2015, at 16:48, Akihiro Motoki <amotoki at gmail.com> wrote:
> 
> For Horizon, the fix is now in the gate queue.
> https://review.openstack.org/#/c/218786/
> It only skips the related test, but the failing test is specific to vendor plugin implementation
> and we can explore the fix later.
> 
> Regarding the stable releases, Horizon may also need the similar version cap for what we did for Neutron.
> I will check it.
> 
> Akihiro
> 
> 2015-08-31 20:54 GMT+09:00 Sean Dague <sean at dague.net>:
> On 08/31/2015 06:48 AM, Ihar Hrachyshka wrote:
> >> On 31 Aug 2015, at 08:19, Kevin Benton <blak111 at gmail.com> wrote:
> >>
> >> Even if this version is fixed for valid_mac, it appears the netaddr authors made the decision to make a backwards incompatible change WRT to the 'broadcast' attribute on IPNetwork objects that have CIDRs of /31 and /32. This means that all future versions of netaddr will be incompatible with the current releases of Neutron.
> >>
> >> I have a fix for master here: https://review.openstack.org/#/c/218723/, but we will need to cap netaddr in global requirements for kilo and juno and then consider back-porting the changes. Additionally, we should probably release a note indicating that upgrading netaddr is disastrous for all released versions of openstack using Neutron.
> >>
> >> Cheers
> >
> > For Juno, we already cap the version of the library to <= 0.7.13. As for kilo, I pushed the following change: https://review.openstack.org/#/c/218805/
> 
> This looks like it's tanking Horizon unit tests as well -
> http://logs.openstack.org/00/207800/1/gate/gate-horizon-python27/980e8a1/console.html#_2015-08-31_11_36_16_378
> 
>         -Sean


Note that in Neutron, we need to backport the real fix into stable/kilo [1] before master to pass grenade job. Otherwise, Neutron itself may be fine without capping the version in global requirements, but if horizon folks think they don’t want to cherry-pick their fix to stable/kilo, then I am fine to proceed with the cap.

[1]: https://review.openstack.org/#/c/218852/

Ihar
-------------- next part --------------
A non-text attachment was scrubbed...
Name: signature.asc
Type: application/pgp-signature
Size: 455 bytes
Desc: Message signed with OpenPGP using GPGMail
URL: <http://lists.openstack.org/pipermail/openstack-dev/attachments/20150831/88a578e0/attachment.pgp>


More information about the OpenStack-dev mailing list