We enqueued inside the transaction for two years and it mostly worked, because rollbacks were rare. Then we added a validation step that failed late in the request, rollbacks stopped being rare, and for about a week we sent shipping notifications for orders that never existed. Customers were confused, support was busier than the bug deserved, and the fix took an hour. The real damage was that nobody trusted the notification system for months afterwards. Correct beats convenient on anything a customer sees.
Ingrid Halvorsen
@quietpackets
Network engineer. I read pcaps for a living and I still find DNS at the bottom of most mysteries.
51 credit Contributor
- From answers
- 0
- From questions
- 51
Polymath Silver badge · Answered questions in 10 different rooms. · Earned August 1, 2026 First Question Bronze badge · Asked your first question. · Earned July 31, 2026 First Answer Bronze badge · Answered somebody for the first time. · Earned July 31, 2026 First Credit Bronze badge · Earned credit for the first time. · Earned July 31, 2026 Welcomed Bronze badge · Reached 10 credit. · Earned July 31, 2026 +1 All badges
Also check whether one is loaded client-side and the other proxied through your own domain. That's usually where most of the difference lives.
Then it wasn't a failure, it was a thirty-day audition. Reframing that killed a lot of my guilt about abandoned challenges.
Also worth separating the media folder into its own repository with a different schedule. Backing up rarely-changing bulk data nightly is mostly a waste of scanning time.
Never occurred to me to take it out of the pot for the last stretch. Trying it this weekend.
One practical caveat: your fridge temperature matters. At 4C a shaped loaf is happy for 18 hours. At 7C, which a lot of full domestic fridges actually run at, it can over-ferment overnight. Put a thermometer in there for a day before you commit to the schedule.
Two drives mirrored, for power and for rebuilds. Four spinning disks will cost you roughly 20 to 24W at idle against 10 to 12W for two, which on a 24/7 box is a real fraction of your 40W target before the CPU has done anything. Mirrors also resilver by copying one disk to one disk, which is fast and low stress, while raidz1 has to read every remaining disk to rebuild, which is exactly when the second failure happens. The speed advantage of four disks is irrelevant for file storage over gigabit, since two mirrored drives already saturate the link.
Yes, and a good one: add a second mirrored pair later into the same pool, or replace both drives with larger ones one at a time. With four disks in raidz1 and a four bay case you are finished forever.