[14:27:23] <wm-bb>	 <mahir256> Are rate limits when calling Wikifunctions functions from Toolforge/WMCS likely to differ from those when calling them from outside Wikimedia servers?
[14:44:16] <wm-bb>	 <vrandecic> That's a great question, a good idea, and we haven't yet made any decisions. We should add this to T282922 or T282921. Do you want to?
[15:07:01] <wm-bb>	 <lucaswerkmeister> I could also imagine having a separate function orchestrator+evaluator in WMCS, with less availability guarantees but higher rate limits
[15:07:26] <wm-bb>	 <lucaswerkmeister> for some reason I want to say “like the cloud replica DBs” even though it’s not a great comparison – it’s not like people can access the production DBs ^^
[15:07:49] <wm-bb>	 <vrandecic> True. There is no need for having only one orchestrator.
[15:10:11] <bd808>	 A dedicated orchestrator feels like a better idea for scaling than treating traffic from the WMCS origin differently than public traffic. The SREs have solid reasons for not wanting to make special configuration for WMCS traffic into the production server cluster.
[15:11:55] <bd808>	 Something similar to the https://wikitech.wikimedia.org/wiki/Help:CirrusSearch_elasticsearch_replicas cluster would probably not be too difficult to get hardware for in a future budget cycle.
[15:14:57] <wm-bb>	 <Jandit> Is there somewhere an overview of needed components/services for Abstract Wikipedia? Aside of Wikifunctions there seems to be the need to somehow store the "abstract sentence" and some information about the languages it can be rendered to (and probably a lot of other things I have no clue about)