akosiaris added a comment.
For what is worth, we now have the services proxy (envoy based) with persistent connections and doing TLS on its own so any costs from switching to TLS connections to the internal LVS services will be largely mitigated. In fact, if anything I expect the latencies from that part of the equation to decrease since it won't have to go through a proxy and the edge caches. The question of whether bypassing the edge caches will hugely increase the load on mediawiki still stands, but there have been many changes on the mediawiki caching infrastructure too (e.g. we now have onhost memcached) so that might very well be largely mitigated as well. I think we ought to revisit this indeed. Having the updater go through an extra 4 (outgoing proxy + 3 layers of edge caches) layers of the infrastructure, one of which is in NO WAY deemed critical to have High Availability (the outgoing proxy) doesn't help with either easy debugging nor ease of operations during maintenance/emergencies. TASK DETAIL https://phabricator.wikimedia.org/T199219 EMAIL PREFERENCES https://phabricator.wikimedia.org/settings/panel/emailpreferences/ To: akosiaris Cc: Ladsgroup, akosiaris, BBlack, Aklapper, Smalyshev, Gehel, MPhamWMF, CBogen, Akuckartz, Nandana, Namenlos314, Jony, Lahi, Gq86, Lucas_Werkmeister_WMDE, GoranSMilovanovic, QZanden, EBjune, merbst, LawExplorer, Vali.matei, _jensen, rosalieper, Scott_WUaS, Jonas, Xmlizer, jkroll, Wikidata-bugs, Jdouglas, aude, Tobias1984, Manybubbles, Mbch331
_______________________________________________ Wikidata-bugs mailing list Wikidata-bugs@lists.wikimedia.org https://lists.wikimedia.org/mailman/listinfo/wikidata-bugs