This is a fair understanding, but limited
I have been using ZFS on Solaris for a quote a while. I think most things/techniques I know will essentially be ported over, but there may be a couple of changes to the code base. If anyone uses ZFS on a BSD (if there is currently a port), their experience might be a little closer.
Essentially, you are correct. Depending on when you added the extra disk, depending on how you set the zpool (or whatever they are going to call it in OS X) up, depending on what kind of error it truly is . . . you are correct. Your data does not magically become raided just be using ZFS. You have to set it up that way, which means you have to have the space for it. I think for casual users, the idea would still be to have separate backups. It is just that a lot of features (like quick expansion of the "disk" or "volume" your current data is on) is now possible. Indeed, a quick decrease is now pretty easy to (like if you want to do rolling upgrades of storage drives; you add in a couple, expanded the pool to include them, remove the older drives--all while the system is up and data integrity is maintained). I could see people quickly upgrading secondary drives this way without extended system downtime. In Solaris, it is not yet possible to do this with boot drives, though. Even if it will be the default filesystem in Leopard, it does not mean this feature can be implemented (it seems really tough, though I have only casually looked at the ZFS code).
Some of you are asking what is new--well, there is a lot that is new. It may actually take some time for people to get used to the new lingo considering some of the advanced file system concepts coming down the line, but things like disk, filesystem, volumes will be needed some clarification/re-definition. Some advanced journaling filesystem/volume managers (like JFS2 on AIX) have been able to do the active growing thing--even on boot volumes--but it is harder to decrease the size (because of the way filesystems used to reserve space). Some filesystems have had snapshots (come in real handy for accidentally and recently deleted files). Some have had built-in raid (all kinds of different options). I am not aware of anything out there that has all these features, and are so easily packaged and administered. I am definitely not aware of anything that is free (Veritas' stuff might come close, though I am not sure they have native snapshotting on VxFS yet), and I am aware of nothing that even comes close for the Mac.
Elaborate, please?I think that previous poster had a valid point to some extent... if you have an intelligently spanned ZFS array with more than one drive, and one of them fails while it has data on it, you must either have the ZFS configured so that data is fully redundant on another drive within the system, or the backup is kept on a separate set of drives. I don't see any other way to get around the issue of how to recover if you have two drives pooled together and one of the drives fails. My understanding is ZFS deals with this, but not in a scenario like the one that poster mentioned, where you have just two drives of different sizes and there's no way to achieve full physical redundancy.
I have been using ZFS on Solaris for a quote a while. I think most things/techniques I know will essentially be ported over, but there may be a couple of changes to the code base. If anyone uses ZFS on a BSD (if there is currently a port), their experience might be a little closer.
Essentially, you are correct. Depending on when you added the extra disk, depending on how you set the zpool (or whatever they are going to call it in OS X) up, depending on what kind of error it truly is . . . you are correct. Your data does not magically become raided just be using ZFS. You have to set it up that way, which means you have to have the space for it. I think for casual users, the idea would still be to have separate backups. It is just that a lot of features (like quick expansion of the "disk" or "volume" your current data is on) is now possible. Indeed, a quick decrease is now pretty easy to (like if you want to do rolling upgrades of storage drives; you add in a couple, expanded the pool to include them, remove the older drives--all while the system is up and data integrity is maintained). I could see people quickly upgrading secondary drives this way without extended system downtime. In Solaris, it is not yet possible to do this with boot drives, though. Even if it will be the default filesystem in Leopard, it does not mean this feature can be implemented (it seems really tough, though I have only casually looked at the ZFS code).
Some of you are asking what is new--well, there is a lot that is new. It may actually take some time for people to get used to the new lingo considering some of the advanced file system concepts coming down the line, but things like disk, filesystem, volumes will be needed some clarification/re-definition. Some advanced journaling filesystem/volume managers (like JFS2 on AIX) have been able to do the active growing thing--even on boot volumes--but it is harder to decrease the size (because of the way filesystems used to reserve space). Some filesystems have had snapshots (come in real handy for accidentally and recently deleted files). Some have had built-in raid (all kinds of different options). I am not aware of anything out there that has all these features, and are so easily packaged and administered. I am definitely not aware of anything that is free (Veritas' stuff might come close, though I am not sure they have native snapshotting on VxFS yet), and I am aware of nothing that even comes close for the Mac.
