3 ms·
The only problem with atomic filesystem snapshots is the time required to perform recovery which can be avoided. The recovery process is of course reliable.
by mulander 5y ago
The only problem with atomic filesystem snapshots is the time required to perform recovery which can be avoided. The recovery process is of course reliable.
- CodesInChaos 5y agoDoes recovery take a long time? Do you have any numbers? I'd have expected recovery to take time proportional to the size of recent/uncommitted transactions, which should be quick, even for large databases.
- mulander 5y agoI don't have experience with large SQLite databases. I had PostgreSQL take over 40 minutes for recovery on a 2 TB database set to checkpoint every 30 minutes. How long it takes will depend on the WAL size, the frequency of checkpoints and the operations that were performed before a crash/snapshot.