How much is lost?
The bad write
On Wednesday at 14:20, a bad release overwrites players' gardens. Shard Garden's RPO is 1 hour: the team has promised to lose no more than an hour of players' work.
Three protections could bring the gardens back.
Try it
Meet the RPO
Each protection restores to a different moment. Tap one to see what players would lose.
Quick check
Why does a daily backup miss Priya's 1-hour RPO?
- Up to a whole day of changes can come after the last backup
- Daily backups take longer to restore
- Daily backups are not real backups
Show the answer
Up to a whole day of changes can come after the last backup
Anything players did after the last daily backup is lost, and that can be almost 24 hours.
Restore beside, then switch
A Firestore restore writes to a new database. Remy checks the restored gardens there first, then points the game at it. The damaged database stays untouched as evidence.
Quick check
Where does a Firestore restore put the data?
- In a new database, beside the damaged one
- Over the damaged database
- On players' phones
Show the answer
In a new database, beside the damaged one
Restoring beside means a wrong restore can't make things worse, and the damage can still be studied.
Both, not one
PITR reaches back 7 days, a minute at a time. Daily backups kept for 14 weeks reach much further, a day at a time. Together they cover a mistake from an hour ago and one found a month later.
Pick the pair
Which pair covers both a bad write an hour ago and a bug found three weeks later?
- PITR, plus daily backups kept for 14 weeks
- PITR alone
- Weekly backups alone
Show the answer
PITR, plus daily backups kept for 14 weeks
PITR meets the 1-hour RPO for recent mistakes. The long-kept backups reach the older ones PITR can't.
On the studio board
The studio adds to the board:
PITR is on for every Shard Garden database, with daily backups kept for 14 weeks: an RPO of 1 hour. A restore into staging is timed every month against an RTO of 4 hours.
Bolt says: Next: breaking things on purpose.
Fact-checked 12 October 2026
- AWS Well-Architected: Disaster recovery objectives docs.aws.amazon.com
- Cloud Firestore: Point-in-time recovery firebase.google.com
- Cloud Firestore: Back up and restore data firebase.google.com