rdo2: periodic ovb job (master) isn't handling hashes correctly

[Matt Young]

In a related bug: https://bugs.launchpad.net/tripleo/+bug/1773569

I noticed that for the master ovb periodic job running in rdo2:master (which has never passed) we're not setting the hash correctly when we invoke quickstart.

# the ansible-playbook invocation is here:

https://thirdparty.logs.rdoproject.org/jenkins-periodic-master-rdo_trunk-ovb-minimal-pacemaker-multiple-nics-192/console.txt.gz#_2018-05-26_18_36_24_690

# here's the cleaned up call (to avoid eye bleed)

export wd="/home/rhos-ci/jenkins/workspace/periodic-master-rdo_trunk-ovb-minimal-pacemaker-multiple-nics"

ansible-playbook -vv $wd/playbooks/baremetal-full-deploy.yml
-e @$wd/config/release/master.yml
-e @$wd/config/nodes/1ctlr_1comp.yml
-e @$wd/config/general_config/ovb-minimal-pacemaker-multiple-nics.yml
-e @$wd/config/environments/ci-rhos.yml
-e local_working_dir=$wd
-e virthost=localhost
-e @$wd/config/environments/ovb-common.yml
-e cleanup_stacks_keypairs=false
-e dlrn_hash_tag=
-e overcloud_image_url=http://REDACTED/ci-images/centos-org-image-cache/master//overcloud-full.tar
-e ipa_image_url=http://REDACTED/ci-images/centos-org-image-cache/master//ironic-python-agent.tar
-t all,teardown-nodes

Note the missing params "//" on the image URL's, it should be "/$hashValue/". Also "dlrn_hash_tag" is empty.

Once we get past the blocking issue linked above, this is what we'll run headlong into.

Capturing it here so we can track/fix.

[Quique Llorente (quiquell)]

The problem is here https://thirdparty.logs.rdoproject.org/jenkins-periodic-master-rdo_trunk-ovb-minimal-pacemaker-multiple-nics-192/console.txt.gz#_2018-05-26_18_35_28_357

get_delorean_hash_from_url is not resolving it.

[Quique Llorente (quiquell)]

Job is missing "current_build" if you check other jobs the environment variable is there:

https://rhos-dev-jenkins.rhev-ci-vms.eng.rdu2.redhat.com/job/periodic-queens-rdo_trunk-featureset020-1ctlr_1comp_64gb/85/

[Matt Young (halcyondude)]

Posted a patch to resolve this...

https://codeng/gerrit/#/c/141919

It looks like we've been appending all these months following a patch in october that changed a tee --> tee -a, the plugin finally gave up :)

[Matt Young (halcyondude)]

update, the job also had a trigger put there by hand that was causing it to be run outside the context of the wrapping multijob. since JJB updates wierdly don't remove existing triggers when they update a job, it was firing daily without params...which is why some of the jobs had no current_build. The patch posted was not the root cause (though it was another issue worth fixing).

[Matt Young (halcyondude)]

I removed the trigger (by hand) and confirmed that JJB pushes don't bring it back, as well as confirmed by inspection that there is no trigger in place in JJB / source control.