- backup
- ebs snapshots, no snaplock?
- mongodump (its not great?)
- maybe only run ebs on a snapshot node
- provisioning
- from a snapshot: good
- secondary sync: alright
- mongorestore: not great
- periodic secondary sync
- compacts and repairs collections and dbs
- hard on your primary
- resets padding factor for all collections to 1
- RAID
- deciding volume count is hard later, decide early
- use piops
- mongodb::backups
- cron job on backup host, NOT PRIMARY
- ec2-consistent-snapshot of :mongo_volumes
- locks db
- arbiters
- processes that only vote
- snapshots
- often
- backup node should be priority 0, hidden 1
- if you've got extra arbiters remove backup node from voting
- run continuous compaction
- fragmentation
- underuse of memory
- deletes not the only source
- fix by resync, repair, or compaction
- db.touch to warm up secondaries?
- mms monitoring
- 10gen hosted solution
- single python script
- pretty