puppet-openstack-unit-6.11-* fails because of os_woekers_* tests failing

[Takashi Kajinami]

puppet-openstack-unit-6.11-ubuntu-bionic and puppet-openstack-unit-6.11-centos-8 are now failing because of failures in test cases for os_workers.

Example:
 https://zuul.opendev.org/t/openstack/build/de5d64c70e0d47689636fa1726e7d518
 https://zuul.opendev.org/t/openstack/build/a8e899e6f98b4ac7881ea3518f8a61e1

[Takashi Kajinami (kajinamit)]

Seems that we observe the failure after recent bump of facter package from 2.5.7 to 4.0.12 .

We didn't see these falure when facter 2.5.7 was used.
 Reference https://storage.bhs.cloud.ovh.net/v1/AUTH_dcaab5e32b234d56b626f72581e3644c/zuul_opendev_logs_611/712770/1/check/puppet-openstack-unit-6.11-centos-8/611fa90/job-output.txt

[Damien Ciabrini (dciabrin)]

For what I can see locally, it seems that facter doesn't report fact processorcount (even though it reports fact processors) as it used to do with e.g. 3.14:

✗ bundle exec facter --version
4.0.11

✗ bundle exec facter processors
{
  count => 8,
  isa => "x86_64",
  models => [
    "Intel(R) Core(TM) i7-8650U CPU @ 1.90GHz",
    "Intel(R) Core(TM) i7-8650U CPU @ 1.90GHz",
    "Intel(R) Core(TM) i7-8650U CPU @ 1.90GHz",
    "Intel(R) Core(TM) i7-8650U CPU @ 1.90GHz",
    "Intel(R) Core(TM) i7-8650U CPU @ 1.90GHz",
    "Intel(R) Core(TM) i7-8650U CPU @ 1.90GHz",
    "Intel(R) Core(TM) i7-8650U CPU @ 1.90GHz",
    "Intel(R) Core(TM) i7-8650U CPU @ 1.90GHz"
  ],
  physicalcount => 1
}

✗ bundle exec facter processorcount

[root@buche ~]# facter --version
3.14.2
[root@buche ~]# facter processorcount
8

[Damien Ciabrini (dciabrin)]

After more digging, I think this is most likely due to the fact that we're pulling facter 4.0.x [1], and I think there's a typo for the EL implementation of fact processorcount, which is called processorscount instead [2].

[1] https://github.com/puppetlabs/facter-ng
[2] https://github.com/puppetlabs/facter-ng/blob/master/lib/facts/el/processors/count.rb#L8

[Alex Schultz (alex-schultz)]

If it was just the el version, the ubuntu jobs shouldn't fail for the same reason

[Alex Schultz (alex-schultz)]

CI also gets facter 2.5.7

[Alex Schultz (alex-schultz)]

I filed a bug against facter but we'll see. https://github.com/puppetlabs/facter-ng/pull/400

[Takashi Kajinami (kajinamit)]

I'm still waiting for ci result to confirm this, but guess the gate failure itself was solved by https://review.opendev.org/#/c/714469/.

I notice the following errors in unit test outputs before we merged the patch, which can be caused by some change in Puppet ?

2020-03-22 13:49:56.653347 | centos-8 | os_workers_heat_engine
2020-03-22 13:49:56.653385 | centos-8 | with processorcount=1
2020-03-22 13:49:56.659488 | centos-8 | [2020-03-22 13:49:56.659404 ] ERROR Facter - --flush-- not implemented but required