We monitor client side errors using a basic mechanism described in this blog post. and monitored here. Occassionally we need to dive into the numbers. This explains how.
Detect trends
editTurnilo can be used to get a sense of abnormal trends. For example an error might be specific to the iOS browser.
Last hour:
By user agent If no trends are prevalent it's likely a generic bug.
Smoke test
editIf a trend is detected in a certain browser see if you can replicate it by performing smoke tests of common functions e.g. most errors relate to iPhone X go test on one either a real device or browserstack.com !
Stat1004
editTo do this I look at recent referrers with the problem using hive on stat1005.eqiad.wmnet. Sometimes clues such as pages, refererrer and projects can help narrow down where the error is occurring visit the page incognito until I see an error in the developer console. The time I use is the current time in UTC.
use wmf;
select uri_host,referer, count(*) as c from webrequest where
day = 7
and month = 5
and year = 2020
and hour = 9
and referer = 'Mozilla/5.0 (iPhone; CPU iPhone OS 13_4_1 like Mac OS X) AppleWebKit/605.1.15 (KHTML, like Gecko) Version/13.1 Mobile/15E148 Safari/604.1'
and uri_path LIKE '%beacon%' and uri_query LIKE "%WebClientError%" group by uri_host,referer sort by c asca