> From: David Lang [mailto:da...@lang.hm] > > > The whole point of journaling is that the filesystem effectively does "fsck" > on the fly, every time it accesses an inode, it checks the consistency. That > way, the work of fsck is spread out during normal operation, rather than > requiring manual intervention, or a really long wait time for system to reboot > after crash. > > this is one of those theory vs practice things (in theory, theory and practice > are the same, in practice they are not) > > in theory a journaled filesystem never needs to be checked. > > in practice it's not always true. It's almost always true that the filesystem > will be usable after an unexpected shutdown, but usable != clean.
I agree, except, that "usable != clean" is irrelevant. The whole point of the journal is that your filesystem doesn't *need* to be clean. Any part of the FS that isn't clean, by definition, you are not using. As soon as you use it, it becomes clean. Taking it a step further, btrfs and zfs, by design, cannot ever become inconsistent. Like journaling on steroids. That's not to say it's impossible to lose data - It's only to say, that the sum total of data risk is that which is buffered in ram, waiting to be flushed to disk at the time of crash. But by design, at all times, the entire filesystem on disk is guaranteed to be a fully consistent snapshot of the whole filesystem, at a point in time, free of any inconsistencies. _______________________________________________ Tech mailing list Tech@lists.lopsa.org https://lists.lopsa.org/cgi-bin/mailman/listinfo/tech This list provided by the League of Professional System Administrators http://lopsa.org/
