overcloud deploy blocked. Can't deploy overcloud due to missing NTP parameter in the overcloud-deploy.sh script

[Francesco Pantano]

BUG DESCRIPTION
===
I'm using latest master of tripleo-quickstart, this implies containerized undercloud.
Running the quickstart (not featureset related), with an arbitrary number of nodes which of course involve 3 controllers, when the overcloud-deploy role is reached, it fails because the ntp_server parameter is missing unless user define it (or override it) in a custom_env yaml file that should be passed to the quickstart execution.

Since the operator can override this value providing the ntp_server variable, the error is caused if the user doesn't specify this parameter in an env.yaml because a default value is missed, causing the following error during the *overcloud deploy* execution:

https://pastebin.com/hLfLCk5S

In a scenario with `controllers > 1` this value is relevant and the overcloud-deploy role in the controller sections has a couple of tasks for configuring correctly chrony (as described here https://docs.openstack.org/install-guide/environment-ntp.html), so this value is expected to be found.

Further, this value is used by the pre-deploy task to set a fact which will be used by the pre_deploy to build the overcloud-deploy.sh script (a j2 template).

This cause the stop with an error of the the overcloud-deploy playbook (that has a task to build this bash script), and the operator has to patch manually that script to deploy the overcloud.

How to reproduce the error:
--

Using quickstart just deploy an overcloud with 3ctrl_1comp or 3ctrl_1comp_3ceph
using featureset049 for the ceph deployment scenario, feature001 for the rest; of course, you can use your preferred featureset, it is not relevant.

Some considerations
---

This parameter is featureset independent, it make sense in a deployment scenario in which services need to be synchronized (it means controllers > 1), so this parameter should go on all the node files affected by this issue.
The second consideration is related to the default that we should assign to the
variable.
As done for the undercloud, the default could be pool.ntp.org, if some scenarios need a change (i.e. install behind a proxy), the operator can ovveride the parameter.

SOME TEST CASES TO CONFIRM THE ISSUE:
==
1. quickstart with 3ctrl_1comp + featureset001 => Verified
2. quickstart with 3ctrl_2comp_3ceph + featureset049 => Verified
3. quickstart with 3ctrl_ + featureset001 => Verified
4. quickstart with 1ctrl_1comp + featureset001 => IT DOESN'T HAPPEN AS EXPECTED
5. quickstart with 3ctrl_3comp + any featureset => VERIFIED

[OpenStack Infra (hudson-openstack)]

Fix proposed to branch: master
Review: https://review.openstack.org/634267

[wes hayutin (weshayutin)]

need more information re: this issue. Please provide the full command line invocation

[Francesco Pantano (fmount)]

Hi @wes,
the full command line invocation could be one of the list I've written above, the only constraints you have is to apply for nodes with controller > 1.
For instance, you can run:

./quickstart.sh -R master -T all -S tripleoui-validate -c ~/.quickstart/config/general_config/featureset049.yml -N ~/.quickstart/config/nodes/3ctlr_2comp_3ceph.yml -E ~/mycustomparameters.yml

As a result, running the overcloud-deploy.sh, you can see the error described in the "Description" section.
My idea of put the ntp_server variable in the nodes files couldn't be the best solution from a style point of view, but I suggest to have at least a default in a file that users are forced to pass to the quickstart (just to have a clean execution).

[OpenStack Infra (hudson-openstack)]

Change abandoned by Francesco Pantano (<email address hidden>) on branch: master
Review: https://review.openstack.org/634267