Hi Taavi,

Thanks for this awesome writeup, and your contributions to the stack!

I wanted to share that recent polling and consensus gathering within WMF's Product Engineering unit also mirrors your insights. Beta Cluster / development rig challenges surfaced as the most important engineering priority area where Product Engineering is interested to help.

There seems to be shared sentiment here, as we've also observed signals in developer satisfaction surveys [1].

Coincidentally, we're aiming to assemble something of a strike force to *begin* its work *sometime* (TBD) in the Wikimedia Foundation's Q3 (January 2023 - March 2023). The idea is to spend 2-3 months initially to see how much traction we can get with an initial plan. Product Engineering management and some tech leads will be meeting over the course of the next several weeks to determine approach for enlisting Product Engineering talent in addressing issues, in concert with our colleagues in the WMF Technology department and elsewhere.

We'd be very interested to collaborate. I'll follow up off thread to see if there may be ways we can better connect.

Thanks again! And I'd also like to express appreciation for the cool things going on with mwcli and Patchdemo I think we're seeing some pretty good hints at ways we might gain better assurance about reproducibility with less effort expended in daily engineering life.

Heads up, I'm on break next week to spend time with family, and will be catching up on my inbox upon return.

Respectfully,

Adam Baso (he/him/his/Adam) Director of Engineering Product Engineering Wikimedia Foundation

[1] cf. https://www.mediawiki.org/wiki/Developer_Satisfaction_Survey/2021/Trends

I also see your reference to the Beta Cluster survey.

On Fri, Nov 18, 2022 at 8:58 AM Taavi Väänänen hi@taavi.wtf wrote:

Hi all,

As you might know, the Beta Cluster (aka deployment-prep)[0] hasn't had an official maintainer for a while, and as a result of that its infrastructure has accumulated a large amount of technical debt.

The Beta cluster currently tries to emulate the production MediaWiki cluster, but it doesn't have any official support from the SRE team[1][2]. The SRE team, on the other hand, has very different needs from a staging cluster than what Beta currently offers and what the MediaWiki developers need.

I wrote a longer blog post on the subject[3], but I believe it's reasonable to conclude that we should start thinking on what might be used to provide replacements for Beta's current use cases, instead of thinking on how we can keep the current Beta alive in the long term.

It's going to be a fairly large project which would need dedicated resources, but so would keeping Beta up to date with infrastructure changes (for example the MW-on-K8s project which will completely overhaul how MW is deployed and served).

Taavi

---

Wikitech-l mailing list -- wikitech-l@lists.wikimedia.org To unsubscribe send an email to wikitech-l-leave@lists.wikimedia.org https://lists.wikimedia.org/postorius/lists/wikitech-l.lists.wikimedia.org/