[09:23:31] <phedenskog>	 I'm adding AlertManager to all front end alerts today. It will probably generate some new emails but it will be the same alerts as before.
[10:10:15] <gilles>	 https://web.dev/yahoo-japan-news/
[10:37:56] <phedenskog>	 I was always thinking high CLS caused more clicks per session :)
[10:49:56] <phedenskog>	 You still need to enable Chrome to show the "fast page" link right?  I tried and I need to enable to show it. The Swedish Wikipedia was marked as fast.
[13:57:52] <phedenskog>	 It looks like the "fast page" mark is a manual or semi manual thing. Tested web sites that have all green in crux but they miss the fast page.
[15:34:08] <gilles>	 it's per group of URLs, so some Swedish Wikipedia URLs might be marked as fast, others might not
[15:34:26] <gilles>	 the feature might be enabled by default in some areas
[15:35:05] <gilles>	 same for the websites you tested, maybe their front page is good in CRuX but other URLs that were grouped with the main page were not
[15:35:19] <gilles>	 how they group URLs is a secret sauce
[15:35:49] <gilles>	 I think it's not just URI format, it's also the structure of the HTML
[19:21:21] <phedenskog>	 AaronSchulz: I added AlertManager alerts for savetimings today but it seems that one of the graphs broke when I saved it: https://grafana.wikimedia.org/d/000000362/save-timing-alerts?viewPanel=2&orgId=1 "Frontend Save Timing". I couldn't see anything in the diff but the dashboard was saved one year ago so maybe Grafana did something? Or that data is from coal, is that broken?
[19:21:51] <phedenskog>	 I can revert the changes to the version before.
[19:24:44] <phedenskog>	 Let me revert.
[19:26:06] <phedenskog>	 No that didn't help.
[19:26:18] <phedenskog>	 https://usercontent.irccloud-cdn.com/file/iexMJxCR/Screenshot%202021-03-10%20at%2020.25.59.png
[19:27:29] <phedenskog>	 I'll have a look first thing tomorrow. I restored back to my latest changes.
[19:28:22] <phedenskog>	 Is there any reason we use coal in that graph and mw.performance.save.p75 in the p75? 
[19:28:43] <phedenskog>	 instead of mw.performance.save.p50?
[20:03:20] <AaronSchulz>	 seems like it could all use the later
[20:04:12] <Krinkle>	 phedenskog: afaik we alwyas use p75 and p95 for RUM data. we haven't used p50 for a long time
[20:04:14] <Krinkle>	 ignores a lot of data
[20:04:30] <Krinkle>	 except for some things with low sample rate
[20:05:09] <Krinkle>	 but if they are that low, it probably  isn't a good thing to track anyway for real users, so even for country dashboards we use p75
[20:05:22] <Krinkle>	 at p50 they would be almost just as jiggly
[20:05:32] <Krinkle>	 (eg. < 20 data points per minute)