- Documentation, documentation, documentation
- More flexibility: if possible per-table CSVs
- More flexibility: Physical backups
- Better recovery documentation: "one line to recover" (there is now a recover_section.py)
- Faster point in time recovery/premade tools
- Better compression
- Prepare based on name of the backup, not just the section
- Option optimization (e.g. double the use_memory)
- More detailed health checks of backups (size, failures, objects, ...). E.g. check size is within a percentage of the previous backup.
- Identify failures after X amount of timeout/time passed and easy cleanup of file leftover (probably on T205627)
- Purge old metadata and make sure logs are rotated T205627 T349360
- Review and improve logging (beyond metadata)
- 1 retry after initial failure
- More optimization of certain database tables
- Maybe some kind of locking of backups and/or transfer.py to prevent concurrent actions on the same source or target servers
- Incremental/Differential backups T244884
- Document the last edit time (and potentially alert on) of some sample tables (e.g. recentchanges or revision) to verify the source databases are up to date (e.g. if its master, or intermediate master have replication stopped, or some other issue causing recent backups of stale data)
- Have a quick way to see which backup sources belong to each section (tendril, dashboard) (Done on puppet's site.pp
- Document and/or automate best server configuration for fast dump load (e.g. disable checksums, innodb transactionality, etc.)
- Enable the possibility of editing per-table options such as the engine and compression
- Workaround the "myloader doesn't import empty dbs" bug
- Binlog backups
- Easier/automated recovery/backup testing
Related incident: https://wikitech.wikimedia.org/wiki/Incident_documentation/20160623-etherpad