Meta Stack Exchange is a question and answer site for meta-discussion of the Stack Exchange family of Q&A websites. It only takes a minute to sign up.
Sign up to join this community
The user activity overview dialog that you see when you mouse over your user name at the top of the page normally displays the reputation that you gained on the current day. This was handy for telling if you were close to the reputation cap.
I just noticed that the dialog is now displaying the combined reputation that I gained yesterday and today (3.5 hours into a new day).
Was this an intentional change to display the reputation gained in the last 24 hours, or is this a display/caching bug?
We run a job once a day to reset these totals to 0 for everyone exactly at UTC midnight, unfortunately when deploying network-wide changes to fix associations permanently some servers went absolutely batshit crazy due to a few race conditions. Our status dashboard looked like a Christmas tree for the better part of a hour. Hopefully though, you saw very few errors during this.
The reputation dropdown is driven off of 5 columns in the database from your user record: Reputation, ReputationWeek, ReputationMonth, ReputationQuarter and ReputationYear. The way we maintain these is incredibly simple, a trigger. When Reputation changes, that same delta is applied to all the other columns. When the interval rolls over, the job running at UTC midnight resets whatever intervals are changing to 0 for everyone in a simple update.
Due to the server farm FREAKING THE HELL OUT, that didn't happen. Honestly, it's just not worth fixing IMO, since it only affects the day interval and will fix itself tonight in the rollover.
If it's that important to people we can fix the totals for today, but unless there's a large cry to do so, I'll spend time on bigger fish today (e.g. showing actual rep change from when a user is deleted in a sensible way, or post migrations as "migrated" instead of "removed" in the rep history).
Init() code on StackAuth/API v1 was running many, many many times due the timeout/retry mechanism in the pipeline - producing some very bad side-effects. Once we sorted out that memory gobbling (literally all of it) problem, everything settled down.
Mar 16, 2012 at 0:31