The initial view and repeat view of the Videos page isn't cached: http://www.webpagetest.org/result/120313_DC_3JRTT/ First view: 9.720s Repeat view: 9.311s So mkelly is looking into adding django-cache-machine to dev, so we can test.
(By "Videos" page, I mean https://firefoxflicks-dev.allizom.org/en-US/recent).
Uh, 9 seconds? Unless we're committing some crimes there, even an uncached request shouldn't take so long...
Yeah, this isn't meant to actually fix the 9 second load time, but since our videos aren't cached at all right now it should help at least a bit. I'll be talking with jd tomorrow about how to find the real cause of the slowness.
Commit pushed to develop at https://github.com/mozilla/firefox-flicks https://github.com/mozilla/firefox-flicks/commit/522981e51677b350852d2ad4f573a1018dd9ddd7 Hook videos and UserProfiles to cache. fix bug 735498
Status: NEW → RESOLVED
Last Resolved: 6 years ago
Resolution: --- → FIXED
So, just a few (non-scientific) datapoints (on the /recent page, at least): Prod: http://www.webpagetest.org/result/120314_6M_3JT35/ First View (5.871s) Repeat View (3.469s) Trunk/dev: http://www.webpagetest.org/result/120314_DG_3JT3B/ First View (9.908s) Repeat View (6.375s) (We'd have to repeat this multiple times in order to have a decent baseline, I feel.) There are multiple factors at play, here, one of which is the Zeus/Riverbed caching (or are we on Citrix Netscaler, now?), the other is--likely--memcached--if we're on the generic memcached cluster. I don't know enough about our setup for this on either prod or trunk/-dev, so it feels inconclusive. Also, prod is likely more "warmed up" in terms of traffic/hits. Needless to say, this is proven technology (and, thus, safe), and manual testing doesn't bear any fallout. We'll chase this up again tomorrow, especially with JD's help.
Commit pushed to develop at https://github.com/mozilla/firefox-flicks https://github.com/mozilla/firefox-flicks/commit/2f5303079fc17129ffae60bdd4bd8d6ad3fbccad Hook videos and UserProfiles to cache. fix bug 735498
You need to log in before you can comment on or make changes to this bug.