THT dependency on ansible-role-chrony not met

Bug #1801090 reported by Daniel Farrell
6
This bug affects 1 person
Affects Status Importance Assigned to Milestone
tripleo
Invalid
Low
Alex Schultz

Bug Description

Description
===========

Undercloud deployment is failing on task "Run the package installation script" because openstack-tripleo-heat-templates has a new dependency on ansible-role-chrony that isn't met.

Steps to reproduce
==================

* Download a reproducer_quickstart.sh script from a job, make ZUUL_CHANGES var null so it deploys master. Example: https://pastebin.com/4ECF2vZZ
* Install openstack CLI client.
* Do initial RDO Cloud config so can use that as infra (can provide details but this is basically irrelevant)
* Execute the reproducer script: `bash -x reproducer-quickstart.sh --create-virtualenv true`
* Connect to resulting box with command printed at end: `ssh -o StrictHostKeyChecking=no -o UserKnownHostsFile=/dev/null zuul@38.145.34.237`
* Source vars: `source /home/zuul/env_vars_to_src.sh`
* Execute tests: `/opt/stack/tripleo-ci/toci_gate_test.sh 2>&1 | tee console.log`

Expected result
===============

Deployment and tests pass with no errors.

Actual result
=============

Deployment fails with:

TASK [undercloud-setup : Run the package installation script] *************************************************************************************
Thursday 01 November 2018 11:38:47 +0000 (0:00:03.123) 0:00:40.421 *****
fatal: [undercloud]: FAILED! => {"changed": true, "cmd": "set -o pipefail && /home/zuul/install_packages.sh 2>&1 | awk '{ print strftime(\"%Y-%m-%d
 %H:%M:%S |\"), $0; fflush(); }' > /home/zuul/install_packages.sh.log", "delta": "0:00:03.600200", "end": "2018-11-01 11:38:52.100265", "msg": "non
-zero return code", "rc": 1, "start": "2018-11-01 11:38:48.500065", "stderr": "", "stderr_lines": [], "stdout": "", "stdout_lines": []}

The script in question is doing:

sudo yum install -y python-tripleoclient ceph-ansible

And the failure, just doing that pkg install, is:

Error: Package: openstack-tripleo-heat-templates-10.0.1-0.20181101014554.92ffd5c.el7.noarch (delorean-current)
           Requires: ansible-role-chrony

This patch seems to be what introduced the new dependency:

https://review.rdoproject.org/r/gitweb?p=openstack/tripleo-heat-templates-distgit.git;a=commit;h=771dd5a6b7b8a13c486b097f3cdca04cf076a294

Environment
===========
1. Exact version of OpenStack you are running.

Latest master.

2. Which storage type did you use?

Not sure, but pretty sure not relevant.

3. Which networking type did you use?

Didn't make it this far, but I eventually wanted to use OpenDaylight.

Logs & Configs
==============

install_packages.sh.log: https://pastebin.com/XNgUTcW4
repo_setup.log: https://pastebin.com/p65EETRU
console.log: https://pastebin.com/mRUCVLzE

Revision history for this message
Alex Schultz (alex-schultz) wrote :

You are using an outdated promoted repository. It is available from the current-tripleo repository.

Changed in tripleo:
milestone: none → stein-2
status: New → Triaged
importance: Undecided → Low
assignee: nobody → Alex Schultz (alex-schultz)
Revision history for this message
Daniel Farrell (dfarrell07) wrote :

Thanks for the feedback. Now trying to understand why my seemingly-stock reproducer env is different from what I guess is happening upstream.

Revision history for this message
Alex Schultz (alex-schultz) wrote :

The reproducer for an older job may not have the promoted content. If you rechecked the job and it's still failing then you'll likely get a version of dlrn hash that has the dependency.

Revision history for this message
Daniel Farrell (dfarrell07) wrote :

If anyone else finds this later, the bit I didn't understand was that there's a dlrn_hash_tag variable hard-coded into reproducer-quickstart.sh scripts. Dlrn is a continuous packaging tool that constantly builds RPMs from the latest OpenStack source. The hash represents the version of those RPMs that will be deployed. If it becomes old, you might see problems like this where what's happening in upstream CI is different from what you're deploying yourself. To fix, just find a newer reproducer-quickstart.sh script and use the latest dlrn_hash_tag variable.

My expected behavior was that if I set the ZUUL_CHANGES var to null, the latest artifacts (including RPMs via DLRN) would be used with no additional under-dev changes applied.

Changed in tripleo:
status: Triaged → Invalid
To post a comment you must log in.
This report contains Public information  
Everyone can see this information.

Other bug subscribers

Remote bug watches

Bug watches keep track of this bug in other bug trackers.