Jump to content
MediaWiki

Developer Satisfaction Survey/2026/Deployment

From mediawiki.org

v Β· d Β· e

πŸ“– Developer Satisfaction 2026 Report

πŸš‚ Deployment

[edit ]

tl;dr

  • SpiderPig – 94% of users are satisfied/0% are dissatisfied.
    • SpiderPig has been unreasonably successful in its first full year of operation
    • 41% increase in unique deployers, year-over-year
    • 59% increase in number of changes backported, year-over-year
  • Deployment satisfaction – 66% of respondents said deployments are easy and pain-free. Pretty much flat.
  • Kubernetes satisfaction – 62% of respondents said they were satisfied with using Wikimedia Kubernetes down a touch from last year.


# Deployment use in the past year

"In the past year, have you used the Wikimedia deployment train or backport windows?"

Of the 56 respondents who opted in to the Deployment section of the survey and answered this question:

  • 46% said they used both
  • 25% said they used backport windows only
  • 18% said they used deployment train only
  • 11% said that did not use Wikimedia deployment train or backport windows in the past year

Of the 50 respondents who said they did use Wikimedia deployment train or backport windows in the past year:

  • 52% said they used both
  • 28% said they used backport windows only
  • 20% said they used deployment train only

The majority of MediaWiki Core developers, Mediawiki Extension developers, test engineers, tool developers, and gadget developers who responded said they use both deployment train and backport windows.

top


# Deployment satisfaction

How much would you agree with the following statement: "Deployments using the Wikimedia train or backport windows are easy and pain-free."
last year
  • A majority (66%) said they agreed that these deployments are easy and pain-free, similar to last year.
  • 10% said they neither agreed nor disagreed
  • 22% said they disagreed

top


# Other deployment feedback

"Please share any other feedback you may have about the Wikimedia deployment train and backport windows"
last year

Lack of deployers during backport

25% of responses lament the lack of people able to help with their deployments during the backport window.

There are not enough deployers to cover all backport windows 100% of the time. I have repeatedly scheduled changes for deployment, shown up on IRC before the backport window and waited for it to begin, and even 20 minutes after the scheduled time there are still no deployers available, despite that at least three are listed on the deployment schedule as being available to deploy.

The availability of deployers during the backport windows for volunteer developers, like me, is essential. In my experience of scheduling patches for deployment in the past year, I have faced several instances where no deployers were available to help me deploy my patch.

top


# Kubernetes use in the past year

"In the past year, have you deployed software to Wikimedia’s production Kubernetes infrastructure?"
last year

Of the 106 respondents who opted into the Deployment section of the survey and answered this question, the majority (65%) said they did not deploy software to Wikimedia’s Kubernetes in the past year; 25% said they had; and 10% were unsure.

top


# Kubernetes satisfaction

"How satisfied are you with software deployment using Wikimedia’s production Kubernetes infrastructure?"
last year

Of the survey takers who had used Wikimedia’s production Kubernetes infrastructure during the past year,

  • a majority (62%) of respondents said they were satisfied with using Wikimedia Kubernetes, down from 70% last year
  • 19% were neither satisfied nor dissatisfied
  • 19% were dissatisfied

top


# Kubernetes ease of deployment

How much would you agree with the following statement: "Deployments using Wikimedia’s Kubernetes infrastructure are easy and pain-free?"* *Question excludes those who reported not deploying to Wikimedia Kubernetes
last year
  • 50% agreed that deployments were easy and pain-free
  • 12% neither agreed nor disagreed (down from 25% last year)
  • 35% disagreed (up from 13% last year)
  • 3% were unsure

top


# SpiderPig use

"In the past year, have you used Wikimedia’s SpiderPig deployment tool?" last year

Of the 48 respondents who opted in to the Deployment section and who answered this question, the majority (58%) said they did not use SpiderPig in the past year; 38% said they did use SpiderPig in the past year; and 4% said they weren't sure.


top


# SpiderPig satisfaction

"How satisfied are you with Wikimedia’s SpiderPig deployment tool?" last year

Of the 18 respondents who said they did use SpiderPig in the past year,

  • 94% said they were satisfied
  • 6% said they were neither satisfied or dissatisfied
  • 0% said they were dissatisfied


top


# Other Kubernetes feedback

"Please share any other feedback you may have about Wikimedia’s Kubernetes production infrastructure"
last year

Debugging/Ease of use

Overall, a small pool of open answers to this question. 33% of them mentioned something related to debugging.

It doesn't quite feel natural, eg I feel like I'm following a recipe carefully as I do service deployments, but it works painlessly and it seems clear enough how to fix a downgrade if necessary (it hasn't been yet). Overall this is a HUGE improvement from the olden days of manual service setup but not quite as easy as SpiderPig.

AltStyle γ«γ‚ˆγ£γ¦ε€‰ζ›γ•γ‚ŒγŸγƒšγƒΌγ‚Έ (->γ‚ͺγƒͺγ‚ΈγƒŠγƒ«) /