Open Stack

On 10/28/2015 9:22 AM, Matt Riedemann wrote:
>
>
> On 10/28/2015 9:06 AM, Yuriy Nesenenko wrote:
>> Hi. Look at https://review.openstack.org/#/c/239837/
>>
>> On Wed, Oct 28, 2015 at 3:52 PM, Matt Riedemann
>> <mriedem at linux.vnet.ibm.com <mailto:mriedem at linux.vnet.ibm.com>> wrote:
>>
>>     That job is failing at a decent rate, tracking with bug:
>>
>>     https://bugs.launchpad.net/cinder/+bug/1510656
>>
>>     It lines up with the novaclient 2.33 release on 10/27, I'm checking
>>     out what the change was that caused the regression.
>>
>>     This is a heads up that rechecks on this failure probably won't help.
>>
>>     So far I haven't seen any related patches up to fix it although
>>     there were already 2 bugs reported when I got in this morning.
>>
>>     --
>>
>>     Thanks,
>>
>>     Matt Riedemann
>>
>>
>>
>> __________________________________________________________________________
>>
>>     OpenStack Development Mailing List (not for usage questions)
>>     Unsubscribe:
>>     OpenStack-dev-request at lists.openstack.org?subject:unsubscribe
>>
>> <http://OpenStack-dev-request@lists.openstack.org?subject:unsubscribe>
>>     http://lists.openstack.org/cgi-bin/mailman/listinfo/openstack-dev
>>
>>
>>
>>
>> __________________________________________________________________________
>>
>> OpenStack Development Mailing List (not for usage questions)
>> Unsubscribe:
>> OpenStack-dev-request at lists.openstack.org?subject:unsubscribe
>> http://lists.openstack.org/cgi-bin/mailman/listinfo/openstack-dev
>>
>
> Heh, well that's 3 bugs then, I didn't see that one. jgriffith and I
> were talking in IRC about just handling both exceptions in cinder to fix
> this but we also agreed that this is a backward incompatible change on
> the novaclient side, which was also discussed in the original novaclient
> wishlist bug that prompted the breaking change.
>
> Given the backward compat issues, we might not just be breaking cinder
> here, so I've proposed a revert of the novaclient change with
> justification in the commit message:
>
> https://review.openstack.org/#/c/239941/
>
> At least with the cinder change above we're OK for mitaka, and logstash
> isn't yet showing failures for cinder in stable/liberty, but given the
> requirements there it will be a failure in cinder python34 tests in
> stalbe/liberty also - so we can backport the cinder fix or block the
> 2.33 novaclient version on stable/liberty global-requirements depending
> on what we do with the proposed novaclient revert.
>

I have an alternative to the revert here:

https://review.openstack.org/#/c/239963/

That makes novaclient.exceptions.RequestTimeout extend requests.Timeout 
so that older cinder continues to work.

I also have changes to block novaclient 2.33.0 in g-r on master and 
stable/liberty:

https://review.openstack.org/#/q/I6e7657b60308b30eed89b269810c1f37cce43063,n,z

I personally think we need to block 2.33.0 since it breaks cinder, then 
release a new version of novaclient with either the revert or the 
alternative change to extend requests.Timeout.

If we block novaclient 2.33.0 then we can also revert the workaround in 
cinder (which would start breaking if we reverted the new exception type 
out of novaclient w/o blacklisting 2.33 first).

-- 

Thanks,

Matt Riedemann