EventBus is definitely tier-1. It provides a communication backbone that is used for reliable event processing. Delays in update processing cause directly user-visible issues like outdated information, performance regressions in VisualEditor & other use cases, and unavailability of public event streams. Timely processing of events like revision deletions is security-relevant.

On Fri, Nov 18, 2016 at 9:07 AM, Nuria Ruiz <nuria@wikimedia.org> wrote:

Team:

(services cc-ed)

We need to define whether kafka on the analytics cluster is a tier-1 or tier-2 system in order to decide the level of support. At this time an issue with a kafka host pages the whole ops team and I am not sure whether this is needed.

From the standpoint of 'anything' analytics kafka is a tier-2 piece on infrastructure (https://wikitech.wikimedia.org/wiki/Analytics/Tier2). I am not sure as to the EventBus use cases ,though. Can the services team detail the use cases for EventBus that would make it tier-1?

I want to make sure we are all in the same page when it comes to define something as tier-1: tier-1 means that is as important as the database or varnish or any of the core pieces of infrastructure that allow us to operate. Powering a feature that users use, like, say, a part of the Android application doesn't make a service tier-1, it just makes it user-facing. That does not necessarily imply that it needs 24/7 support.

Could services team let us know as to the clients of Event Bus as of now that would make it a tier-1 service?

Thanks,

Nuria

Gabriel Wicke

Principal Engineer, Wikimedia Foundation