Linux: EXT3

From mmmv_kos
Jump to: navigation, search

Linux : failisüsteemid

Väidetavalt(arhiivkoopia)kirjutab andmed kõigepealt ja siis alles meta-andmete logi, erinevalt EXT4 failisüsteemist, mis pidavat andmeid kirjutama alles pärast meta-andmete logi kirjutamist. Eelnevalt viidatud väite tsitaat:

All modern file systems are power fail tolerant. The only one that is regularly used that isn't is FAT, or ext2 if you still use that.

Now this just means your metadata will be consistent, any data that hasn't been written out to disk yet will be lost. ext3 will write all dirty data before it commits the journal every 5 seconds, so it is least likely to screw you, you will only lose the last 5 seconds worth of data. Ext4, btrfs and xfs all do this thing called delayed allocation, which means we avoid writing data until the last possible moment, so you could very well lose the last 5 minutes of data written but still have a consistent file system if your applications are not using fsync().