Failed to build cradox on CentOS7

[Marcin Dulak]

openstack-ansible 17.0.0.0b2 fails during the setup-infrastructure.yml step "TASK [repo_build : Create OpenStack-Ansible requirement wheels]" with "Failed to build cradox".

[Marcin Dulak (marcin-dulak)]

Logging into the dev-os1-hci2_repo_container-3e0d8cfb container and executing `pip install cradox` manually results in:

Exception: gcc, python-dev, librados2 or librados-dev >= 0.80 are missing

Actually only librados2-devel seems to be missing. Manual installation of `yum install -y librados2-devel` inside of the container results in successful `pip install cradox`.

[root@dev-os1-hci2-repo-container-3e0d8cfb ~]# rpm -q librados2-devel
librados2-devel-0.94.5-2.el7.x86_64

[Laurent Frisée (laurent-f)]

Sorry should have put that in comment, also affect ubuntu 16.04.3.

Same behavior as the last post from Marcin Dulak, apt install librados-dev seems to solve the issue.

Version of librados-dev installed using apt command line
librados-dev 12.2.1-0ubuntu0.17.10.1~cloud0

[Jean-Philippe Evrard (jean-philippe-evrard)]

17.0.0b2 is very old now (for a bleeding edge branch I mean), and if you're willing to use the development branch, maybe you try to run directly master? Maybe requirements have moved since then...

But in any way, that sounds an important bug at first sight: a missing dependency.

Could you help us triage this by doing two things:
- Post your user space configuration (user_variables and openstack_user_config.yml), and what you're trying to achieve.
- git checkout master; git pull origin master; and try to redeploy the whole thing (don't forget to bootstrap ansible again, we bumped the version since b2).

Thank you in advance.

[Marcin Dulak (marcin-dulak)]

I don't experience this problem when starting from a fresh master eb67707b5ddabef8b4536324bc2c096f23f16d17 on CentOS7, and librados2-devel-0.94.5-2.el7.x86_64 is present in the container.

[Jean-Philippe Evrard (jean-philippe-evrard)]

Thanks for the feedback.

Could you tell us what your user config are? I'd like to understand in which context does that bug trigger, because our ceph deploys seem to work fine on master ubuntu.

[Marcin Dulak (marcin-dulak)]

I think I used the attached openstack_user_config.yml, keeping the user_variables.yml intact.

[Jean-Philippe Evrard (jean-philippe-evrard)]

This was solved by this: https://review.openstack.org/#/c/530570/

[Jesse Pretorius (jesse-pretorius)]

This has been fixed with https://review.openstack.org/530570