[07:31:17] godog: I've changed them. [08:18:54] phedenskog: ack! so the 'team' label is now 'perf' in the alerts, correct ? [08:19:12] Yes but I've only changed two. [08:19:29] wanna play around with for a while before I change everything :) [08:20:30] heheh ok, I'll change back the review to use 'perf' [08:21:46] phedenskog: LMK if https://gerrit.wikimedia.org/r/c/operations/puppet/+/663238 looks good to you [08:21:53] if it does I'll merge and we're good to go [08:23:17] I'm looking at your questions in the task too [08:26:28] godog: yes go ahead [08:26:59] ok! will do [08:37:03] phedenskog: I'm not sure I understand the first point of https://phabricator.wikimedia.org/T272979#6825655, or rather the end result you'd like to achieve [12:13:47] godog: ok let me try to explain it better: first thing that I didn't mention in the meeting, I think the long term goal for us it not to act on the performance alerts as the performance team, instead each product should act on them, that's why I want them to be easy to understand. That's like the long term goal. [12:13:47] At the moment each of our testing have three our four conditions that needs to be true for an alert to fire. At the moment each condition is repeated in the alert manager with a metric and a description. So we show the same description four times and then each metric that fired. That's for one alert. Then if we group them and we four alerts that fires at the same under one group, we will have 4*4 metrics/descriptions. I think [12:13:47] that will be heard to read/understand. [12:15:06] For many dashboard each condition doesn't mean anything for me, but the combination of all four conditions fires the alert, so in best case scenario I only need one description to say what's happening, instead of repeating the same over and over. [12:15:55] That's why I'm thinking there's a way to convert the alert from Grafana when it arrives in alert manager to make it easier to understand. [13:12:04] phedenskog: thank you! now I get it I think, yeah I'm not aware of any such way to (if I got it right) essentially express another condition on the received alerts [13:13:30] phedenskog: although perhaps inhibition of alerts could work in your case? https://prometheus.io/docs/alerting/latest/alertmanager/#inhibition [20:05:51] question for you folks: during the visual diff tests comparing parsoid vs core rendering, i sometimes get noisy diffs because of a perf popup "did the page load fast enough"? I want to hide that popup via js .. can you point me to the HTML / js code that renders it so i can figure out what query selectors to use in the visual diffing preprocessing code? [20:29:22] subbu: This is powered through the QuickSurveys extension afaik [20:29:36] ok. [20:29:38] it probably/hopefully has a central mechanism to turn it off or opt-out, maybe a cookie or query parameter [20:29:52] if not, I can come up with a css rule to hide it [20:32:19] if a query param, i would probably add it to https://github.com/wikimedia/integration-visualdiff/blob/master/bin/examples/parsoid.php.diffsettings.js#L21 but css rule also works. something i can add to https://github.com/wikimedia/integration-visualdiff/blob/master/lib/php_parser.postprocess.js ... will poke around later today and if necessary will ping you. [20:34:43] I believe ?quicksurvey=(any string not matching "true" or "*-survey-*") would turn off surveys completley, e.g. ?survey=none should do. I could not find what typical value is used for off but "0" or "none" or "off" etc would do [20:35:08] also, I see a CSS rule here that probably works to turn it off: https://github.com/wikimedia/mediawiki-extensions-QuickSurveys/blob/63cfe627449d1efd8da02b0f683a1e20a49cb577/resources/ext.quicksurveys.lib/styles.less#L79 [20:35:26] I don't know if `.ext-quick-survey-panel ` completely covers it, but might work [20:36:12] oh, actually, the query param won't work because NavTming perf survey doesnt use internal sampling logic, we bypass that [20:37:34] yeah, so probably need css then [20:38:57] https://ru.wikipedia.org/wiki/%D0%91%D0%BE%D1%84%D0%BE%D1%80%D1%82,_%D0%94%D0%B6%D0%BE%D0%B0%D0%BD?quicksurvey=internal-survey-perceived-performance-survey [20:39:02] yeeh, that css selector will cover it