[opensuse] raid 1 with btrfs
I said: * one ssd (450Gb) for main disk (30GB system, btrfs, 380GB data, xfs) I want to add some explanation. When I speak of "data" in the line above, I mean user data (home, for example), not data in the BTRFS meaning. I can manage to take some time to recover user data in case of crash. But if BTRFS/RAID have to be useful, I need a booting computer. If the crash is not that of the disk but the computer itself, I may not be able to rebuild a similar hardware, so the question: is the BTRFS raid 1 mirror single disk readable (better bootable) on an other computer? mdamdm disk are not, not without a small but non obvious change thanks jdd -- http://dodin.org -- To unsubscribe, e-mail: opensuse+unsubscribe@opensuse.org To contact the owner, e-mail: opensuse+owner@opensuse.org
On 10/02/18 16:52, jdd@dodin.org wrote:
I said: * one ssd (450Gb) for main disk (30GB system, btrfs, 380GB data, xfs)
I want to add some explanation. When I speak of "data" in the line above, I mean user data (home, for example), not data in the BTRFS meaning.
I can manage to take some time to recover user data in case of crash. But if BTRFS/RAID have to be useful, I need a booting computer.
If the crash is not that of the disk but the computer itself, I may not be able to rebuild a similar hardware, so the question: is the BTRFS raid 1 mirror single disk readable (better bootable) on an other computer?
You mean if you lost one drive of two? If it's a true mirror, quite likely.
mdamdm disk are not, not without a small but non obvious change
What small change? provided you've set the md mirror up correctly, you can take a single disk, stick it in a new machine, and boot. The only problem is the boot may "fail" with "can't mount root" initially until you force the array to assemble degraded. At which point you can then add a new disk and rebuild the array. I suspect that btrfs is just the same - with a disk missing the partition won't start up properly. Fix that and you should be able to boot off a single disk. Cheers, Wol -- To unsubscribe, e-mail: opensuse+unsubscribe@opensuse.org To contact the owner, e-mail: opensuse+owner@opensuse.org
Le 10/02/2018 à 19:36, Wol's lists a écrit :
If the crash is not that of the disk but the computer itself, I may not be able to rebuild a similar hardware, so the question: is the BTRFS raid 1 mirror single disk readable (better bootable) on an other computer?
You mean if you lost one drive of two? If it's a true mirror, quite likely.
or none, if the computer (mobo, for example) fails, yoiu have two disks but no system
mdamdm disk are not, not without a small but non obvious change
What small change? provided you've set the md mirror up correctly, you can take a single disk, stick it in a new machine, and boot. The only problem is the boot may "fail" with "can't mount root" initially until you force the array to assemble degraded. At which point you can then add a new disk and rebuild the array.
this is what I mean as "non obvious". I can't connect it on an usb dock and read
I suspect that btrfs is just the same - with a disk missing the partition won't start up properly.
booting is not likely as the hardware has changed on this case, but reading it would be great. It's the "on the fly" part that makes me think that the disk may be readable thanks jdd -- http://dodin.org -- To unsubscribe, e-mail: opensuse+unsubscribe@opensuse.org To contact the owner, e-mail: opensuse+owner@opensuse.org
On 10/02/18 19:04, jdd@dodin.org wrote:
What small change? provided you've set the md mirror up correctly, you can take a single disk, stick it in a new machine, and boot. The only problem is the boot may "fail" with "can't mount root" initially until you force the array to assemble degraded. At which point you can then add a new disk and rebuild the array.
this is what I mean as "non obvious". I can't connect it on an usb dock and read
md won't assemble a broken array unless it was already broken when it was shut down. So if a disk fails, you shut down the system, and move the remaining disk elsewhere, then it will have been shut down in a broken state and I believe it will happily reassemble. If the mobo fails and you stick the drive in another machine, then yes you have to manually assemble --force, but that's a safety feature - don't let the user bring up a compromised system without knowing anything about it. (Okay, it should probably do what btrfs does - if the previous boot was okay and the current boot isn't, it should fail, but it is what it is. The btrfs boot seems to be a not-that-popular compromise either :-) At the end of the day, if the system is in a broken state, then you cannot expect clean functional behaviour. Different systems break (or not) in different ways. Cheers, Wol -- To unsubscribe, e-mail: opensuse+unsubscribe@opensuse.org To contact the owner, e-mail: opensuse+owner@opensuse.org
10.02.2018 22:35, Wols Lists пишет:
On 10/02/18 19:04, jdd@dodin.org wrote:
What small change? provided you've set the md mirror up correctly, you can take a single disk, stick it in a new machine, and boot. The only problem is the boot may "fail" with "can't mount root" initially until you force the array to assemble degraded. At which point you can then add a new disk and rebuild the array.
this is what I mean as "non obvious". I can't connect it on an usb dock and read
md won't assemble a broken array unless it was already broken when it was shut down.
It will (or rather, during boot sequence it will wait for some time and then force degraded array to be started). "It" does not mean "kernel md driver" but rather all other components, including user level ones, that come with mdadm package. -- To unsubscribe, e-mail: opensuse+unsubscribe@opensuse.org To contact the owner, e-mail: opensuse+owner@opensuse.org
On 11/02/18 06:01, Andrei Borzenkov wrote:
10.02.2018 22:35, Wols Lists пишет:
On 10/02/18 19:04, jdd@dodin.org wrote:
What small change? provided you've set the md mirror up correctly, you can take a single disk, stick it in a new machine, and boot. The only problem is the boot may "fail" with "can't mount root" initially until you force the array to assemble degraded. At which point you can then add a new disk and rebuild the array.
this is what I mean as "non obvious". I can't connect it on an usb dock and read
md won't assemble a broken array unless it was already broken when it was shut down.
It will (or rather, during boot sequence it will wait for some time and then force degraded array to be started).
"It" does not mean "kernel md driver" but rather all other components, including user level ones, that come with mdadm package.
Well, I think that most of the md guys would consider that a distro (mis)feature. If it's the root fs with a separate home, then it's not too bad - lose the OS and you're cursing but it's recoverable. If it's home ... :-( However, behind all of this is you need to monitor! The system breaking on boot is not going to help you detect a broken raid/btrfs, if the reason the system crashes is raid/btrfs broke underneath it! I put a comment on the wiki about a datacentre with - probably a petabyte array - where a tech just happened to notice that the raid-6 array had two "disk fail" red lights showing... That's what triggered my outburst against the btrfs defaults - a naive user would think adding a second drive is a good idea. If you don't know what you're doing, IT ISN'T! If techies can screw up, what hope do lusers have? Cheers, Wol -- To unsubscribe, e-mail: opensuse+unsubscribe@opensuse.org To contact the owner, e-mail: opensuse+owner@opensuse.org
-----BEGIN PGP SIGNED MESSAGE----- Hash: SHA1 On Saturday, 2018-02-10 at 17:52 +0100, jdd@dodin.org wrote:
I said: * one ssd (450Gb) for main disk (30GB system, btrfs, 380GB data, xfs)
IMO, 30 GB for btrfs system is too small. I would use 80 GB at least. - -- Cheers, Carlos E. R. (from openSUSE 42.3 x86_64 "Malachite" at Telcontar) -----BEGIN PGP SIGNATURE----- Version: GnuPG v2 iEYEARECAAYFAlp/XcYACgkQtTMYHG2NR9ULpACfTUeRUmQDKikSfRF9OOUnnOhH 6OMAnAp6+dfSUGZ6u9xbXK6qtdq3LzwK =gWls -----END PGP SIGNATURE----- -- To unsubscribe, e-mail: opensuse+unsubscribe@opensuse.org To contact the owner, e-mail: opensuse+owner@opensuse.org
On 10/02/18 21:01, Carlos E. R. wrote:
On Saturday, 2018-02-10 at 17:52 +0100, jdd@dodin.org wrote:
I said: * one ssd (450Gb) for main disk (30GB system, btrfs, 380GB data, xfs)
IMO, 30 GB for btrfs system is too small. I would use 80 GB at least.
Or even more. I think my default is 80GB, and yet my mum's system ran out of space during an upgrade. It might have been less than 80, I can't remember what it is, but it's more likely to be *more* rather than less. Cheers, Wol -- To unsubscribe, e-mail: opensuse+unsubscribe@opensuse.org To contact the owner, e-mail: opensuse+owner@opensuse.org
participants (5)
-
Andrei Borzenkov -
Carlos E. R. -
jdd＠dodin.org -
Wol's lists -
Wols Lists