While investigating T255410 we concluded that we should try to make it clearer to people investigating issues if the timeout was due to:
A) failure to get a response from the termbox Kubernetes service in time
B) we actually got the response from termbox saying it was unable to contact Special:EntityData in time.
We intend the issue in situation A should remain an ERROR from the mediawiki TermboxRemoteRenderer but the B) should become only a NOTICE because an ERROR will already have been emitted by the Kubernetes services (A).
Conceptually the RemoteRenderer (Wikibase service calling the SSR - Kubernetes - service) does the correct thing and falls back to the clientside rendered version.
- Error logging is currently happening in both wikibase and termbox SSR service. At most one error due to a timeout should normally be logged (either from termbox or from wikibase but not both)
- link to "failure to get a response from the termbox Kubernetes service in time" error in logstash
- link to error in logstash stemming from "response from termbox saying it was unable to contact Special:EntityData in time" https://logstash.wikimedia.org/goto/a8c6e830d48e9babeff293c1375bfd33 (see the non-bad language responses in here)
- Issues with Termbox not being able to contact or get response in time from Special:EntityData (or Mediawiki API in general) are logged (by MediaWiki) using NOTICE error level