On Sun, Dec 5, 2010 at 3:51 AM, Helmut Hullen <hul...@t-online.de> wrote: > Hallo, cwillu, > > Du meintest am 05.12.10: > >>>> I am not an expert on this by a long shot, but it looks like you >>>> added these two disks in raid0. > >>> I won't hope that this error is related to RAID0, I haven't >>> installed (as far as I know) RAID0. >>> >>> My installation way: >>> >>> (2-TByte-Disk) >>> >>> mkfs.btrfs /dev/sdf2 >>> mount /dev/sdf2 /srv/MM >>> >>> (1.5-TByte-Disk) >>> btrfs device add /dev/sdc3 /srv/MM >>> btrfs filesystem balance /srv/MM >>> >>> (and then waiting about 1 day ...) >>> Especially: no RAID definition. > > [...] > >> If it's not a raid1, and there's multiple devices, it's a raid0 (and >> so available space is the sum of all drives). Your problem however >> is that metadata is raid1 by default (where everything is duplicated >> on separate drives). > > Maybe you're right. But if you're right then I have got the worst of two > worlds. I don't want neither RAID0 nor RAID1, I want a bundle of > different disks (at least partititions) which seem to be one large disk. > And I've hoped btrfs does this job.
That's what raid0 is, and it's actually the best of both worlds: your metadata (which will be less than 5% of the data) is safely duplicated, such that you always have the checksums even with a disk gone, so you can verify that the data that you still have is good, while not wasting space duplicating every little bit of file data, which you may not care about that much, and which you have backed up anyway (right? right?). Larger files will be striped though, which is more incidental to how data is allocated to block groups, but even so, it's generally not as simple as saying "I don't want raid0, I just want a single big disk"; you have to figure out how to do that, and it'll generally result in some raid0-like properties. This will almost certainly become much more tunable in the future, but not every feature that people want is done yet. In fact, most of the really cool user-visible features aren't done yet. Btrfs is still pretty young. >> Adding another device will probably work around this, as will simply >> running a balance operation (possibly, and you may need to free up >> some space first anyway). > > That could lead to the following steps: > > Buy a 3 GByte disk > > btrfs device add /dev/sdxy /srv/MM > btrfs filesystem balance /srv/MM > > 1.5 TByte disk: > btrfs device delete /dev/sdc3 /srv/MM > btrfs filesystem balance /srv/MM > > and then disconnect the 1.5 TByte disk (and hope that now the 2 TByte > disk sets the limits). > No nice way ... No, just run the balance without adding another disk. That will probably work (although it _will_ take a while on a large filesystem). I'm not sure that you understand how this all works though; you might want to re-read the wiki articles (which I believe have been freshened up recently). > Is there a way to avoid this (presumably) RAID mismatch? Yes, you can specify the raid level for each when you make the filesystem (and will eventually be able to do it with existing filesystems). However, as I described above, you really want metadata to be duplicated. Your problem is more of an unfortunate case of everything not being tuned quite right yet. > By the way: working with TByte disks includes (for home users) that > there's no backup ... Not sure why you'd think that. It can't be the bandwidth, and if you can't afford a second drive, there's a good case to be made that you can't afford the data you can't afford to lose. -- To unsubscribe from this list: send the line "unsubscribe linux-btrfs" in the body of a message to majord...@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html
