Hello everyone,

As some of you might have noticed, all traffic serving the wikis has been directed to our MediaWiki on kubernetes clusters since June 18th[1]. While there is still work to be done to complete the transition for all the side pieces, it has been an important milestone for improving our deployment processes and our operational capability and flexibility.

It now comes the time to turn off our legacy bare metal clusters for the website ("appserver") and apis ("api") to complete the transition.

If you don't run services or jobs within the Wikimedia production environment, you can stop reading here! If you are, please read along:

We plan to do a soft turnoff of those old clusters on the 16th of July during the MediaWiki Infrastructure deployment window[2], and then remove everything persistently 24 hours later.

What changes for you if you're running a service or jobs within production? * If your program is using the service mesh to talk to MediaWiki, nothing will change for you, your requests were already directed to MediaWiki on kubernetes * If your program is reaching out to the wikis via their public url, nothing will change for you, as your requests were also already directed to MediaWiki on kubernetes * If your program was connecting directly to either " appservers-{rw,ro}.discovery.wmnet", "appservers.svc.{eqiad,codfw}.wmnet", "api-{rw,ro}.discovery.wmnet", "api.svc.{eqiad,codfw}.wmnet", you should modify it to connect to https://mw-api-int.discovery.wmnet:4446 if you need to do edits/modifications, and https://mw-api-int-ro.discovery.wmnet:4446 otherwise.

Cheers,

Giuseppe [1] https://phabricator.wikimedia.org/T362323#9903729 [2] https://wikitech.wikimedia.org/wiki/Deployments#deploycal-item-20240716T1700