Re: Production readiness reviews

[John Belamaric]

Hi Jed, thanks for your interest. I think a kickoff meeting is the next step.

I'd like to consider whether we want this to be a one-off effort to introduce the policy and related tooling, or if we see a need for an ongoing health/maturity subproject that would incorporate this as well as items like alpha/beta/GA criteria (https://github.com/kubernetes/community/issues/4000). Or a WG.

Given that I'd like to meet before next week's sig-arch meeting, in case we want to bring any discussion to the larger group. So let's try to meet on Friday this week or Monday or Tuesday next week.

Please everyone that is interested in participating enter available times here:

https://www.when2meet.com/?8207054-8Fl3T

Agenda:

 - Intros

 - Define the scope and organization of the effort

 - Discuss tooling ideas for managing prod readiness reviews

John

On Fri, Sep 27, 2019 at 4:10 PM 'Jed Salazar' via kubernetes-sig-architecture <kubernetes-si...@googlegroups.com> wrote:

Hi John,

I'm interested in helping out with this. I'm a Xoogler with some experience with PRRs and have had similar ideas related to reliability and would love to help. 

Where are you at with this effort? What do you think the next steps will be? 

On Thursday, August 1, 2019 at 12:23:14 PM UTC-6, John Belamaric wrote:

Hi SIG-arch,

Internally we have been discussing how we can improve the safety and reliability of Kubernetes in general, and in upgrades and the use of new features in particular. We thought it may be time to introduce production readiness reviews for Kubernetes features, and would like feedback on this idea.

The goal here is to ensure that new versions and features can be rolled out to clusters in a safe, controlled manner, and can be monitored, debugged, and disabled if necessary. This starter KEP provides a bit more detail:

https://github.com/kubernetes/enhancements/pull/1193/files

Please feel free to comment here or, preferably, on the PR.

While this does add some process overhead, I think that at this stage in Kubernetes' development, it's critical that upgrades do not cause outages for workloads (or for the control plane as much as possible). Without the kind of supports described here, that will be impossible to achieve.

John

--
You received this message because you are subscribed to the Google Groups "kubernetes-sig-architecture" group.
To unsubscribe from this group and stop receiving emails from it, send an email to kubernetes-sig-arch...@googlegroups.com.
To view this discussion on the web visit https://groups.google.com/d/msgid/kubernetes-sig-architecture/5d005f18-b3d7-48d2-bb22-5ce3db328adf%40googlegroups.com.