Looks like somewhere new oslo.config module is used with old configuration options. I'm not sure how this is possible. Could be some serialization/deserialization code that has constructed config options from old data and thus bypassing normal construction and lacking some new object properties.

Unfortunately, I do not have cycles currently to look into this in more detail.

On Mon, Sep 22, 2014 at 2:40 PM, Ihar Hrachyshka <ihrachys@redhat.com> wrote:
-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA512

(Fixing the wrong Doug's email address.)

On 22/09/14 11:23, Alan Pevec wrote:
> 2014-09-22 10:52 GMT+02:00 Ihar Hrachyshka <ihrachys@redhat.com>:
>> Havana jobs failed, probably due to new oslo.config release.
>> Since tags for the last release are already created, we won't
>> fix those failures.
>
> That's correct, 2013.2.4 tags have been pushed and we're not fixing
> this in Havana but it should be investigated. Oslo releases are
> supposed to be backward compatible so we need to understand if
> that's an oslo.config regression or not.
>

I suspect this may be due to oslo-config-options-validation work done
in 1.4.0 [1]. That said, the failure is observed in glance/havana
only, and logs are not very informative. The only thing that is
present is the following critical error in glance-api log (no traceback):

2014-09-22 06:27:04.281 13352 CRITICAL glance [-] 'StrOpt' object has
no attribute 'type'

[1]:
https://review.openstack.org/gitweb?p=openstack%2Foslo.config.git;a=commitdiff;h=2422d4118c97734067ea0b37ae159bc2e3c492c5

/Ihar
-----BEGIN PGP SIGNATURE-----
Version: GnuPG/MacGPG2 v2.0.22 (Darwin)

iQEcBAEBCgAGBQJUH/yhAAoJEC5aWaUY1u57WDoIAMMYBpio1k0iYVoQyFAi5BdI
306BXUsywij1DBdTdbEwqiRET2yAmFPuD4IaYXX0aeVu2quHvRxvzGCLRZQa6jXS
I5ICFlOzBdCBqdauVXzqiipNr7ADSTIJ/QLexPLG3JbEsJ+F1LaFx+SsO+oI0J86
64MbbjORpr0nF5PNp5MaDbrCLvlnYyJEag76Am9UQsqTnS6YYsFyCLD45uBd223T
c3x5+gPuw/20qh2dakcOoe242JcINJbNuWbDUC81mf99NAHuS0xvwatlll+rW2xS
o9aKFB3sCK8J92GyHY/vDdiZGrRbsLgzLGjYo5teULSgV8b5qkcbF36egFd/5LM=
=dOJu
-----END PGP SIGNATURE-----