[ruby-core:113224] [Ruby master Bug#19596] Decreased performance after upgrading from ruby 2.7.2 to ruby 3.2.2
From:
silva96 via ruby-core <ruby-core@...>
Date:
2023-04-13 13:38:51 UTC
List:
ruby-core #113224
Issue #19596 has been updated by silva96 (Benjam=EDn Silva). File clipboard-202304130930-c24f0.png added File clipboard-202304130936-vtuwy.png added File 4511491.json added File clipboard-202304130938-rj5ny.png added byroot (Jean Boussier) wrote in #note-3: > > But I don't see a noticeable impact on the overall response times. So t= hat makes me think that maybe, ruby 3.2 reports differently than 2.7 in new= relic >=20 > Yes, extremely likely. Can you show the whole breakdown and not just the = "ruby" part? If response time stayed similar but "ruby" went up, it means o= ther parts went up. >=20 > I suspect it's simply that you version of New Relic is no longer hooking = properly in one of the libraries you use for IOs (e.g. redis/memcached/db). Here's the whole breakdown for the past 3 days.  Also, here's a profiling session:  (Also attache the whole JSON) And this is the GC Execution I just enabled:  ---------------------------------------- Bug #19596: Decreased performance after upgrading from ruby 2.7.2 to ruby 3= .2.2 https://bugs.ruby-lang.org/issues/19596#change-102767 * Author: silva96 (Benjam=EDn Silva) * Status: Open * Priority: Normal * ruby -v: ruby 3.2.2 (2023-03-30 revision e51014f9c0) +YJIT [x86_64-linux] * Backport: 3.0: UNKNOWN, 3.1: UNKNOWN, 3.2: UNKNOWN ---------------------------------------- I just upgraded to ruby 3.2.2 from 2.7.2 in one Rails app in production.=20 Tried with and without YJIT, but no matter what, I'm having worst performan= ce. Time spent in ruby went from 20ms to more than 50ms and very inconsistent (= peaks) - I'm using RVM - I've set MALLOC_ARENAS_MAX=3D2 I may be wrong, but I suspect I have worst performance because I see this c= hart in newrelic increased right after the upgrade  In the picture, the chart has the "ruby" toggle, to see how much time is sp= ent in ruby operations. But I don't see a noticeable impact on the overall response times. So that = makes me think that maybe, ruby 3.2 reports differently than 2.7 in newrelic  Right now, I'm using YJIT, and I was expecting to see some kind of improvem= ents in app performance upgrading from 2.7 to 3.2 YJIT, but there's no perf= ormance gain, is that even possible? The website is https://profe.social, it's a very standard content site, no = super expensive ruby computations, most operations are rendering erb views = pulling data from the database. Any Ideas how can I get performance improvements? Best regards. ---Files-------------------------------- clipboard-202304130005-opyyk.png (120 KB) clipboard-202304130014-vnb0h.png (134 KB) clipboard-202304130930-c24f0.png (253 KB) clipboard-202304130936-vtuwy.png (618 KB) 4511491.json (1010 KB) clipboard-202304130938-rj5ny.png (100 KB) --=20 https://bugs.ruby-lang.org/ ______________________________________________ ruby-core mailing list -- ruby-core@ml.ruby-lang.org To unsubscribe send an email to ruby-core-leave@ml.ruby-lang.org ruby-core info -- https://ml.ruby-lang.org/mailman3/postorius/lists/ruby-c= ore.ml.ruby-lang.org/