Talk:Admin tools development/SUL Audit/Metrics 20140704

39 millions accounts touched 12–18 months ago is fun! Clearly those merged by the scripts last year.
Apart from bugzilla:61876, «non-attached accounts who do not clash with another [global] account» should approach zero (they were 69) but are 10,596? For these the existing script should
$wgCentralAuthAutoMigrate worked! We have only 250k unattached accounts touched in last 12 months, down from 1.4M last year. Still over 4M total but it didn't get worse. Most interesting will be «who clash with a global account» numbers.
What I find very mysterious is that 50,634... 80 local unattached accounts for each of those 50k global usernames on average?! (We don't have a baseline and bugzilla:61876 needs to be fixed first to have a "real" value.) Does it mean they're all very complex cases with multiple persons involved, or that they were very crosswiki users?

I think the 18 months column might be slightly misleading. I used the AccountAudit values for those who had logged in after it was deployed. Not all users did, so I checked if their user_touched is before the AA deployment. If it was, I used it. Otherwise, I just said they last logged in a second before AA was deployed, since we have no better estimate (I suppose we could guess based on revision/logging, but eh), which is why a bunch of users got lumped into the 18 month column.
Yes, once we run that, it should hit 0.
Just added those numbers in :)
Hm, not sure. The two or three I looked at manually were people missing one or two accounts. I could probably do a breakdown to see how many each account is conflicting with. Legoktm (talk) 20:48, 5 July 2014 (UTC)Reply
- I forgot that of course "older" clashing usernames don't have a corresponding global account, because they should have logged in at least once in last year to get one created automatically. This is now shown very clearly by the "who clash with a global account" row. Which increased only 10k last year. --Nemo 07:06, 6 July 2014 (UTC)Reply