Open Stack

On Fri, Jan 19, 2018 at 4:47 PM, John Trowbridge <trown at redhat.com> wrote:
>
>
> On Fri, Jan 19, 2018 at 10:21 AM, Wesley Hayutin <whayutin at redhat.com>
> wrote:
>>
>> Thanks Marius for sending this out and kicking off a conversation.
>>
>> On Tue, Jan 2, 2018 at 12:56 PM, Marius Cornea <mariusc at redhat.com> wrote:
>>>
>>> Hi everyone and Happy New Year!
>>>
>>> As the migration of tripleo-upgrade repo to the openstack namespace is
>>> now complete I think it's the time to create a Pike branch to capture
>>> the current state so we can use it for Pike testing and keep the
>>> master branch for Queens changes. The update/upgrade steps are
>>> changing between versions and the aim of branching the repo is to keep
>>> the update/upgrade steps clean per branch to avoid using conditionals
>>> based on release. Also tripleo-upgrade should be compatible with
>>> different tools used for deployment(tripleo-quickstart, infrared,
>>> manual deployments) which use different vars for the version release
>>> so in case of using conditionals we would need extra steps to
>>> normalize these variables.
>>
>>
>> I understand the desire to create a branch to protect the work that has
>> been done previously.
>> The interesting thing is that you guys are proposing to use a branched
>> ansible role with
>> a branchless upstream project.  I want to make sure we have enough review
>> so that we don't hit issues
>> in the future.   Maybe that is OK, but I have at least one concern.
>>
>> My conern is about gating the tripleo-upgrade role and it's branches.
>> When tripleo-quickstart is changed
>> which is branchless we will be have to kick off a job for each
>> tripleo-upgrade branch?  That immediately doubles
>> the load on gates.
>
>
> I do not think CI repos should be branched. Even more than the concern Wes
> brought up about a larger gate matrix. Think
> about how much would need to get backported. To start you would just have
> the 2 branches, but eventually you will have 3.
> Likely all 3 will have slight differences in how different pieces of the
> upgrade are called (otherwise why branch), so when
> you need to fix something on all branches the backports have a high
> potential to be non-trivial too.

Once we release we expect the upgrade/update process to be stable and
no changes required to the process so I expect the backports to be
minimal, mostly for scenarios that we missed in testing at release
time.

> Release conditionals are not perfect, but I dont think compatibility is
> really a major issue. Just document how to set the
> release, and the different CI tools that use your role will just have to
> adapt to that.
>>
>>
>> It's extemely important to properly gate this role against the versions of
>> TripleO and OSP.  I see very limited
>> check jobs and gate jobs on tripleo-upgrades atm.  I have only found [1].
>> I think we need to see some external and internal
>> jobs checking and gating this role with comments posted to changes.
>>
>> [1]
>> https://review.rdoproject.org/jenkins/job/gate-tripleo-ci-centos-7-containers-multinode-upgrades-pike/
>>
>>
>>>
>>>
>>> I wanted to bring this topic up for discussion to see if branching is
>>> the proper thing to do here.
>>>
>>> Thanks,
>>> Marius
>>>
>>>
>>> __________________________________________________________________________
>>> OpenStack Development Mailing List (not for usage questions)
>>> Unsubscribe:
>>> OpenStack-dev-request at lists.openstack.org?subject:unsubscribe
>>> http://lists.openstack.org/cgi-bin/mailman/listinfo/openstack-dev
>>
>>
>
>
> __________________________________________________________________________
> OpenStack Development Mailing List (not for usage questions)
> Unsubscribe: OpenStack-dev-request at lists.openstack.org?subject:unsubscribe
> http://lists.openstack.org/cgi-bin/mailman/listinfo/openstack-dev
>