Latest apt-clone upgrade breaks bootup

[loomy]

As of Jan 9, doing an apt-clone on the last version or two of nexentacp causes the bootup to not work. After apt-clone: With one version I got a grub config menu, and with another I got a "enter root password to do maintenance" screen.

[Erast (erast)]

Any chance you could provide full output of apt-clone operation ?
Did something went wrong with upgrade and apt-clone didn't ask you to rollback upgrade?

[loomy (loomy)]

The apt-clone finishes successfully and requests a reboot, so I type reboot. I don't know how to get the full output text out of the VM.

I don't know what you mean by "apt-clone didn't ask you to rollback upgrade?", but when it failed upon reboot and went to the command line, I used the zfs snapshot rollback of blah@blah-003 to restore to the previous snapshot, rebooting worked after that. The second time when it errored to the grub config command line, I didn't know how to fix it and reverted to a VM snapshot.

I expect this is totally reproduceable for you because it worked for me against two different snapshots created by apt-clone. So I had upgraded a few times over the past month or so, and doing apt-clone today against two of those snapshots broke the OS.

[Erast (erast)]

It seems you tried to rollback syspool's snapshot manually instead of using 'apt-clone -r rootfs-nmu-003' command?
Could you confirm that please?

[loomy (loomy)]

The "apt-clone" broke the OS even before doing any rollbacks. When I did do a rollback, I believe I used a syntax similar to 'apt-clone -r rootfs-nmu-001@rootfs-nmu-003'. I wasn't even sure if that was right, but it worked. What didn't work was the "apt-clone" before and also after that :)

[Erast (erast)]

Any chance to reproduce it with freshly installed RC1 ?

[loomy (loomy)]

I can reproduce it in a VM. I have a snapshot of an install that you can reproduce it in if you want to try it out. I can provide a location from which you can download it. I don't think the problem is an immediate concern, but it might be useful for you to see what is broken.

I haven't tried reproducing it from a fresh install.

[Erast (erast)]

I recently found that out-of-space scenario causing syspool:/boot/grub/menu.lst not to be written and as the result GRUB menu is not appearing.. is that similar to that?

In general, to fix this issue, I need steps to reproduce it, i.e. like: 1) install RC2; 2) do this; 3) do that;
Or otherwise it is highly difficult to fix the issue...

[loomy (loomy)]

I don't know anything about solaris, but it sounds like that could be it because I figure there isn't too much broken in this area.

I don't have the exact steps I did to transform the RC1 install into the install I have right now, but I CAN say that all I did was use the apt-clone command. I would have used apt-clone to update, then type the activate command and reboot. I probably did that a couple times between RC1 and the time of the report. I then used apt-clone to remove a snapshot or two. I may have used the zfs command, but I don't think so.

[loomy (loomy)]

PS: I don't know if anyone checks the Q/A section, so maybe you can
comment on this "how to get cifs server working" question which I posted
there: https://answers.launchpad.net/nexenta/+question/22835

Thanks :)

Erast wrote:
> I recently found that out-of-space scenario causing
> syspool:/boot/grub/menu.lst not to be written and as the result GRUB
> menu is not appearing.. is that similar to that?
>
> In general, to fix this issue, I need steps to reproduce it, i.e. like: 1) install RC2; 2) do this; 3) do that;
> Or otherwise it is highly difficult to fix the issue...
>

[Tim Spriggs (tim-tajinc)]

I have seen variations on this where grub isn't updated correctly or zfs file system versions are too old. Is this fixed for the NCP2 upgrade?