[Gluster-infra] Gluster project using CentOS infra for Test/Release work

Fri Feb 12 15:09:15 UTC 2016

Le vendredi 12 février 2016 à 14:13 +0000, Karanbir Singh a écrit :
> Hi,
> 
> Some of us met at Devconf for a conversation around how best the
> CentOS infra might be used within the Gluster projects. Here is a
> quick catchup from that meeting, and to further encourate conversation
> around the topic.
> 
> Meeting attended by:
> - Michael Schearer

That's Scherer, not Schearer.

> - Amye Scavarda
> - Humble Chirammal
> - Kaleb Keithley
> - Niels de Vos
> - Jim Perrim ( CentOS )
> - Karanbir Singh ( CentOS )
> 
> The existing gluster devel workflow was established as :
> - Upstream dev work happens in git
> - Gerrit queues are used to manage the content / contributor flow
> - Jenkins is used as the orchestration tool to validate contributions
>   + Jenkins slaves are deployed against a central Jenkins Server
>   + each Jenkins slave represents a role or distro or test req
>   + Jenkins slaves are run both locally, as well as in
>     RackSpace sponsored Cloud Instances
>   + Packaging is something that may or maynot be tested ( some
>     projects do, others dont )
> - Release Managers own the cutoff / feature specs
>  + Content of each release
>  + Documentation
>  + Releases are not executed in packaged paylogs
> - Packaging team ( or folks from dev/release teams )
>  + Local set of rpms done from gluster.org
>  + Liase with delivery upstreams / distros for packaging in their space
> 
> We established that there are 2 major points of concern at this time.
> 1) Jenkins and Gerrit self hosted are flaky and need unreasonable amount
> of real time effort to keep running
> 2) Being able to test in real world deployment scenarios is hard.
> 
> The CentOS Project curates a build and test infrastructure that is open
> to projects like Gluster,
> who have a SIG presence and have a user story in the CentOS Ecosystem.
> We feel being able to help
> the Gluster project is key for us. Added to that, we can also use the
> Gluster artifacts and expose
> them to related layered projects who have a dependancy on Gluster's
> projects directly ( eg. ovirt,
> openstack etc ), in the same infrastructure.
> 
> Therefore I would like to welcome the Gluster project to come and use
> our Jenkins instance at https://ci.centos.org/ and use the downstream
> testing + release mechanics we have in the project. I also feel we are
> able to get this done with little impact to the larger developer
> expectations in the Gluster projects. Some of the points highlighed:
> - Self hosted Gerrit should move to Gerrithub.io!
> - Are all the needed plugins available on Gerrithub ?
>   + if not, can an easy compromise be found ?
> - Will the existing history migrate to Gerrithub ?

Since we did use the existing history to verify commits for the previous
server compromise, I would stress the importance of keeping the history
here, for security measures. That's kinda not negotiable. 

> - Developer discipline will need to be established to negate the impact
>   of duplicate / redundant identities ( same person multi email add etc )
> 
> In a followup meeting, Brian Stinson ( CentOS Team ) spent
> time with MS ( Gluster QE ) to explain the flow of how
> ci.centos.org is setup, how tests can be depoyed, and results
> collected, notifications run and what the expectations would be.
> 
> There are some highlights from the centos infra : we have an automated,
> baremetal driven test infra. That allows multinode deployments, scale
> out testing. We are also going to have a cloud native test infra in the
> near term, allowing for cloud workload testing - again from a user and
> developer perspective. The resource limits, while they do exist, are
> liberal and in many cases, negotiable. All of this runs as a service to
> you guys, in the gluster project, so you can focus on writing awesome
> storage code, let us run the Jenkins.
> 
> We also established that there are some machines waiting bring up in the
> Community Cage - depending on how the Gluster project can use them or
> not, we would welcome including them in the larger CentOS community side
> pool, ensuring they get used well.
> 
> Brian, Jim and I will be in the Bangalore office on the 3rd
> and 4th of March. Brian has offered to run a demo and face
> to face onboarding process for the Gluster team in the office
> on the 3rd in the afternoon, should this be seen as valuable.
> 
> CentOS infra is deployed, in production now. We can start bringing jobs
> over right away, to suite your requirements / release timelines.

So we could start right away with a new builder.

Could you provision a centos 7 slave, using our slat setup as basis, to
see how it go ?
https://github.com/gluster/gluster.org_salt_states/tree/master/jenkins

For now, centos 7 is what is missing, and as I am busy with the
migration to RH DC, I will not be able to give much help directly for
now.

-- 
Michael Scherer
Sysadmin, Community Infrastructure and Platform, OSAS

-------------- next part --------------
A non-text attachment was scrubbed...
Name: signature.asc
Type: application/pgp-signature
Size: 836 bytes
Desc: This is a digitally signed message part
URL: <http://www.gluster.org/pipermail/gluster-infra/attachments/20160212/aa8d9956/attachment.sig>