On Friday 2018-04-20 15:08, Andrei Borzenkov wrote:

On Fri, Apr 20, 2018 at 2:22 PM, Paul Neuwirth wrote:

...

Hello, web search without X11 is difficult.. Anyone knows how I can boot into openSUSE (Leap 42.3) on a btrfs root filesystem, which got degraded. I wanted to exchange a defective harddisk.

Do you have RAID1? Otherwise what's the point in booting into something that you are going to replace anyway.

I am sorry, it is of course RAID Level 1. One HDD is ok, the other has bad sectors.

...

I tried booting with rootflags=degraded, and also specifying root partition as "root=/dev/sdb2". despite that, boot doesn't even complete to rescue mode, but is waitin without limit with "A start job is running for dev-whatever.device"

Well, if device is not there, then even degraded mount is not possible.

the second device is not needed to mount it degraded. But systemd is awaiting it anyway. Despite of the rootflags.

...

what do I miss?

You did not provide enough details of your configuration to be able to answer it.

The raid is not mdadm raid, but btrfs raid. anything else missing? bootloader is grub2. Regards, Paul -- To unsubscribe, e-mail: opensuse+unsubscribe@opensuse.org To contact the owner, e-mail: opensuse+owner@opensuse.org

20.04.2018 17:31, Paul Neuwirth пишет:

On Friday 2018-04-20 15:08, Andrei Borzenkov wrote:

...

On Fri, Apr 20, 2018 at 2:22 PM, Paul Neuwirth wrote:

...

Hello, web search without X11 is difficult.. Anyone knows how I can boot into openSUSE (Leap 42.3) on a btrfs root filesystem, which got degraded. I wanted to exchange a defective harddisk.

Do you have RAID1? Otherwise what's the point in booting into something that you are going to replace anyway.

I am sorry, it is of course RAID Level 1.

OK, that was expected. As a side note, multi-device btrfs is not really supported by YaST (and installer).

One HDD is ok, the other has bad sectors.

...

...

I tried booting with rootflags=degraded, and also specifying root partition as "root=/dev/sdb2". despite that, boot doesn't even complete to rescue mode, but is waitin without limit with "A start job is running for dev-whatever.device"

Well, if device is not there, then even degraded mount is not possible.

the second device is not needed to mount it degraded. But systemd is awaiting it anyway. Despite of the rootflags.

There are two problems 1. Integration of multi-device btrfs in systemd startup sequence is rudimentary at best. It requires that all devices are seen once and there is no provision for overriding it and booting in degraded state. 2. Such automatic degraded mount would actually be quite dangerous, at least with relatively old kernel used by Leap 42.x. btrfs raid1 gives you exactly one chance to fix it by replacing device. If this single attempt is unsuccessful for whatever reason, filesystem goes in permanent read-only state where you can no more do anything beyond copying data off it and recreating. I believe this is fixed in 4.15. So find yourself live Linux with as recent kernel and btrfs programs as possible and fix your RAID1 from there. -- To unsubscribe, e-mail: opensuse+unsubscribe@opensuse.org To contact the owner, e-mail: opensuse+owner@opensuse.org

On 21 April 2018 at 10:05, Paul Neuwirth wrote:

On Friday 2018-04-20 18:41, Andrei Borzenkov wrote:

...

...

...

...

Hello, web search without X11 is difficult.. Anyone knows how I can boot into openSUSE (Leap 42.3) on a btrfs root filesystem, which got degraded. I wanted to exchange a defective harddisk.

Do you have RAID1? Otherwise what's the point in booting into something that you are going to replace anyway.

I am sorry, it is of course RAID Level 1.

OK, that was expected. As a side note, multi-device btrfs is not really supported by YaST (and installer).

I know, but it was easy, just added another hdd to btrfs fs, and converted to raid 1... just the bootloader module of yast has some trouble with it..

...

...

One HDD is ok, the other has bad sectors.

...

...

I tried booting with rootflags=degraded, and also specifying root partition as "root=/dev/sdb2". despite that, boot doesn't even complete to rescue mode, but is waitin without limit with "A start job is running for dev-whatever.device"

Well, if device is not there, then even degraded mount is not possible.

the second device is not needed to mount it degraded. But systemd is awaiting it anyway. Despite of the rootflags.

There are two problems

1. Integration of multi-device btrfs in systemd startup sequence is rudimentary at best. It requires that all devices are seen once and there is no provision for overriding it and booting in degraded state.

I did not expect it to be, since how long does btrfs raid now exist?

...

2. Such automatic degraded mount would actually be quite dangerous, at least with relatively old kernel used by Leap 42.x. btrfs raid1 gives you exactly one chance to fix it by replacing device. If this single attempt is unsuccessful for whatever reason, filesystem goes in permanent read-only state where you can no more do anything beyond copying data off it and recreating. I believe this is fixed in 4.15.

disgusting.. that's what raid 1 is for fail-safety (and a little performance maybe).. but fortunately data loss is prevented.. I also wondered, why btrfs fs got ro on drive failure.. somehow my case should be a standard scenario in a raid system, without hot-pluggable drives.. hm

...

So find yourself live Linux with as recent kernel and btrfs programs as possible and fix your RAID1 from there.

ok.. so it's maybe better to switch back to mdadm raid and btrfs fs on it. also btrfs raid doesn't have hot spare support.

Thank you for the help.. unfortunately there's no opensuse live system anymore. Is the network installer, booted via PXE, maybe sufficient?

What do you mean there is no openSUSE live system? https://software.opensuse.org/distributions/tumbleweed#livecd-ports is fully supported for this kind of use case https://software.opensuse.org/distributions/testing#livecd-ports will be when Leap 15 is released -- To unsubscribe, e-mail: opensuse+unsubscribe@opensuse.org To contact the owner, e-mail: opensuse+owner@opensuse.org

Paul Neuwirth wrote:

On Saturday 2018-04-21 10:57, Richard Brown wrote:

...

...

Thank you for the help.. unfortunately there's no opensuse live system anymore. Is the network installer, booted via PXE, maybe sufficient?

What do you mean there is no openSUSE live system?

https://software.opensuse.org/distributions/tumbleweed#livecd-ports is fully supported for this kind of use case

https://software.opensuse.org/distributions/testing#livecd-ports will be when Leap 15 is released

I mean the openSUSE install system/shell, there's a install image (initrd + kernel) somewhere repo/oss/boot/x86_64/linux and repo/oss/boot/x86_64/initrd these will do a network setup and will download other data as needed, providing the setup shell... and the setup system will have btrfs, mdadm and so on, i think. and thus this may be sufficient to recreate a clean file system, on a new raid, do chroot and fix other files. that's how I install openSUSE always, not even physical access to the system is needed. a live cd would require that and an optical disk drive.

<pedantic mode> you could load the iso image over pxe too :-) -- Per Jessen, Zürich (19.9°C) http://www.dns24.ch/ - free dynamic DNS, made in Switzerland. -- To unsubscribe, e-mail: opensuse+unsubscribe@opensuse.org To contact the owner, e-mail: opensuse+owner@opensuse.org

Le 21/04/2018 à 10:05, Paul Neuwirth a écrit :

Thank you for the help.. unfortunately there's no opensuse live system anymore. Is the network installer, booted via PXE, maybe sufficient?

yes, leap 15 have many and working good https://software.opensuse.org/distributions/testing jdd -- http://dodin.org -- To unsubscribe, e-mail: opensuse+unsubscribe@opensuse.org To contact the owner, e-mail: opensuse+owner@opensuse.org