[06:43:13] wow, how much can you screw up a cache interface? [06:43:33] * tgr just learned about psr-6 [06:50:35] if you go back to this channel's logs around the time they were voting on it, there's a bunch of commentary on it :P [16:00:29] tgr: I lost all interest in the PSR process when they decided that an HTTP response should be represented by immutable value objects [16:24:51] anomie: your CA caching fix (and tgr's User caching fix) made a pretty big positive dent in save time. See the bottommost graph on https://performance.wikimedia.org/#!/week [16:26:17] ori: What am I seeing there, about 50ms? [16:27:18] I think a bit more, around 80ish. I'll try to quantify it more precisely. [16:32:49] anomie: here's a nice graph: http://graphite.wikimedia.org/render/?width=865&height=556&_salt=1456849910.094&from=-48hours&target=timeShift(movingAverage(mw.performance.save.median%2C100)%2C%221d%22)&target=movingAverage(mw.performance.save.median%2C100) [16:33:54] last 48hrs ; blue line shows the value for the same time on the previous day [16:37:17] without movingAverage() the graph is noisier but it's easier to correlate the change with the sync-files, which went out at 1:03 (for User.php) and 1:40 (for CA): http://graphite.wikimedia.org/render/?width=865&height=556&_salt=1456850150.825&from=-48hours&target=timeShift(mw.performance.save.median%2C%221d%22)&target=mw.performance.save.median [16:38:52] anomie: anyways, I wanted to say thanks -- for sticking with this issue, continuing to follow-up, and tolerating me [16:39:04] ori: No problem. [16:39:06] really appreciate it [22:40:02] tgr, anomie: authmanager and api patches applied on https://authmanager.wmflabs.org/wiki/Main_Page ; UI patch had too many rebase conflicts for me to get it on yet [23:01:07] bd808: the UI patch has not caught up with the latest set of changes to the core patch anyway; I'll get to that soon [23:01:19] *nod*