Re: Openstack-ODL integration - stacking issues with pbr


Kyle Mestery <mestery@...>
 

On Fri, Aug 14, 2015 at 9:56 AM, Flavio Fernandes <ffernand@...> wrote:

On Aug 14, 2015, at 7:00 AM, Ihar Hrachyshka <ihrachys@...> wrote:

-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA256

On 08/14/2015 12:32 PM, Flavio Fernandes wrote:
[cc odl neutron-dev, Ankur, Isaku]


On Aug 14, 2015, at 6:23 AM, Ihar Hrachyshka
<ihrachys@... <mailto:ihrachys@...>> wrote:

On 08/14/2015 12:15 PM, Flavio Fernandes wrote:

On Aug 13, 2015, at 9:04 AM, Ihar Hrachyshka
<ihrachys@... <mailto:ihrachys@...>
<mailto:ihrachys@...>> wrote:

On 08/13/2015 02:46 PM, Flavio Fernandes wrote:

On Aug 13, 2015, at 8:30 AM, Ihar Hrachyshka
<ihrachys@... <mailto:ihrachys@...>
<mailto:ihrachys@...>
<mailto:ihrachys@...>> wrote:

Looking at the logs in the email thread:

2015-08-05 14:59:53.160 | Download error on
https://pypi.python.org/simple/pbr/: [Errno 110]
Connection timed out -- Some packages may not be found!
2015-08-05 15:02:00.392 | Download error on
https://pypi.python.org/simple/: [Errno 110] Connection
timed out -- Some packages may not be found!
2015-08-05 15:02:00.392 | No local packages or download
links found for pbr>=1.3

So why can't your machine download the satisfying pbr
version? It's available on pypi, so assuming you fix
the download error, I think it should proceed.


That is the issue. This happens because OFFLINE=True
and there is still something in pbr requirements that
is looking for a version of pbr that is not used in
stable/kilo.

I suspect this is happening because stable/kilo
branch in networking-odl was created ‘late’ and the
new version of pbr was added as part of liberty?!?

So, to easily reproduce this issue: 1) stack with
devstack+networking-odl on stable/kilo; 2) untack;
3) change OFFLINE=True and 4) attempt to stack
again.


Are your repo [test-]requirements.txt synchronized with
neutron's kilo requirements? They should, otherwise it won't
be ever supported.


They do not for requirements.txt:

$ diff requirements.txt.neutron
requirements.txt.networking-odl | grep pbr <
pbr!=0.7,<1.0,>=0.6
pbr<2.0,>=1.3
$ diff test-requirements.txt.neutron
test-requirements.txt.networking-odl | grep pbr $

So that's a problem. You don't even have a common pbr version that
would satisfy both projects.


indeed!




If you look closely, that is what is changed in the
abandoned gerrit [1]. Maybe the right thing to do is to
re-visit that gerrit and make sure the requirement.txt
files are inline?!?


The fix is obvious: make stable/kilo requirements in your repo
synchronized with what is in stable/kilo for neutron. Same for all
other branches.

Openstack requirements proposal bot can help you maintain the
lists synchronized. To make sure your repo gets updates from the
bot, add it in projects.txt in openstack/requirements repo (in
master and in stable/kilo).


It is ironic reading this… and then looking at the ‘owner' of the
commit that caused all this mess [1].


The bot was proposing when it was still master. After you branched out
stable/kilo, you should have sticked to the deps that belong to
stable/kilo.

That aside; I have very little experience on that… some one
else’s help to take care of this would be greatly appreciated.

Anybody of for that task?


OK, so your repo is in projects.txt, but only on master. That's why
you don't receive updates on stable/kilo. I've cherry-picked the
appropriate openstack/requirements patch to enable the bot for your
stable branch too: https://review.openstack.org/#/c/213084/

I see that you created stable/kilo branch based on some incorrect hash
that already included changes from liberty cycle. That's why you got
the bot requirements update there. You should probably reset your
branch to start from a patch that does not include any Liberty changes.


yes, make sense. How and where is this hash kept? I think whatever hash it
has now is fine, except for the gerrit I propose to be reverted [1]. @Kyle: agree?

One other glitch I see in your repository is that you have stable/juno
and stable/icehouse there while they contain neutron. I guess they
were created when you spinned out your driver out of neutron tree. I
think they can be safely killed now.

Yes. Can you give me more concrete steps on how to accomplish that
‘killing'?


I can get rid of those branches in gerrit, let me do that now.

— flavio




Thanks,

— flavio

[1]: https://review.openstack.org/#/c/197258/



Ihar

-----BEGIN PGP SIGNATURE-----
Version: GnuPG v2

iQEcBAEBCAAGBQJVzcpMAAoJEC5aWaUY1u57Zk8H/3fUBkJVvO/Bad99XGRomh+K
o1ZFpk8wNl5wjdolVJJeJZtYRrF0VaOsowN2Y14s3+8pf6wGMwMTv1nOIazjjcB+
5lUgVGVfcE5A1InxDoy9bH6TgdaLZCGUwAnChBtqmZkmWAAKd1vckWsurVsSJopw
GbTIPsDhncBGDOl4gAYP/ZzO7yzf5JjU6ovsA+KyYRW2Y72ftQf+7XRTUdjWv6Ys
YBuGFaSv+BrEhr5+47nxgwqVWM9PzFuVJkhwaJq5UIPD16zBtG7wCcN3DBjzc+Ki
EI5A8NlJrsHw9slOn9kwA+OY9CnPFMCesVcdY8dz4xSW2rcXAIZgdwDpN5YsiA8=
=9cvr
-----END PGP SIGNATURE-----


Join {z.archive.ovsdb-dev@lists.opendaylight.org to automatically receive all group messages.