Ladsgroup added a comment.
In T246415#6616074 <https://phabricator.wikimedia.org/T246415#6616074>, @Addshore wrote: > In T246415#6541422 <https://phabricator.wikimedia.org/T246415#6541422>, @Ladsgroup wrote: > >> In T246415#6536970 <https://phabricator.wikimedia.org/T246415#6536970>, @Addshore wrote: >> >>> Looking at the ACs I think we have some open questions still >>> >>> - How feasible is it to introduce a "client" and "repo" db load group and have it used in the right places in our code? / how could that be tackled. >> >> That's not easily feasible as lots of queries come from lib/ data-access/ or core itself (e.g. when looking up entity info using page info lookup) and it's pretty hard to find out what's the context this wiki is being ran. > > I guess all this needs though is an abstraction to the DB layer that deals with this and adds the groups where needed? Yup but LB/LB factory is being used directly in lots of places, and if we want to properly inject the abstraction, is going to be really complicated (and pretty big). Some of the core ones doesn't even accept the LB/LBFactory. I'm not saying it's not doable. I'm saying it'll be too much work for too little gain IMHO. > Even lib services ultimately are created by either client or repo? > >> - Having four groups, `client`, `repo`, `repo-terms`, `client-terms`: Not too harder the above but I don't see the point. It wouldn't take advantage of cache locality. > > I don't think we would benefit from having groups such as `repo-terms` or 2 groups joined up for example as this removes flexibility from where DBAs can end up routing these? perhaps? > Might be worth checking with them? Yeah, I'm not sure splitting by both groups would be a good idea (unless we grow really really big.) > The final remaining thing to figure out here would be the AC that reads: > >> [ ] We can move forward from the results of this investigation and implement some db groups that will help the DBAs to split load in a way that will help prevent incidents. > > So we need to answer which direction we want to go in. Seemingly either terms, or client vs repo, or both? I'd say we should collectively decide this. Maybe a meeting would be good? > CC @Ladsgroup @Michael @WMDE-leszek Should this move back to todo? or do we want to try to come up with this final answer not on the camp (but I guess here in stalled/waiting nothing will happen) TASK DETAIL https://phabricator.wikimedia.org/T246415 EMAIL PREFERENCES https://phabricator.wikimedia.org/settings/panel/emailpreferences/ To: Michael, Ladsgroup Cc: ArielGlenn, Michael, Marostegui, Ladsgroup, WMDE-leszek, Aklapper, Addshore, Alter-paule, Beast1978, Un1tY, Akuckartz, Hook696, Iflorez, Kent7301, alaa_wmde, joker88john, CucyNoiD, Nandana, jijiki, Klaas_Z4us_V, Gaboe420, Giuliamocci, Cpaulf30, Lahi, Gq86, Af420, Bsandipan, Pablo-WMDE, GoranSMilovanovic, QZanden, LawExplorer, Lewizho99, Maathavan, elukey, _jensen, rosalieper, Scott_WUaS, Jonas, Wikidata-bugs, aude, Lydia_Pintscher, Mbch331, Jay8g
_______________________________________________ Wikidata-bugs mailing list Wikidata-bugs@lists.wikimedia.org https://lists.wikimedia.org/mailman/listinfo/wikidata-bugs