Bug#925146: grub-probe appears to be stuck (in chroot)

Colin Watson cjwatson at debian.org
Sat Mar 23 13:57:33 GMT 2019


Control: reassign -1 lvm2
Control: forcemerge 918590 -1

On Thu, Mar 21, 2019 at 12:07:34AM +0100, Harald Dunkel wrote:
> I apologize again for not sending the log file you asked for right
> from the start. Hopefully you found the second strace log file?

Indeed - looks like our emails crossed in the post there.

> Of course it is vgs that got slow somehow, but AFAICS it is called
> about 16 times for each grub-mkconfig, even though there are only 4
> logical volumes. Grub-mkconfig was run at least 4 times during the
> upgrade from Stretch to Buster (then I went home, hoping that it is
> finished tomorrow morning).
> 
> Sorry to say, but something got worse with Buster here. If you
> come to the conclusion that it is not a problem of grub, then please
> feel free to reassign this bug report.

Thanks for the details.  I tracked this down and it seems to be pretty
much the same as #918590.  (It's true that grub_util_pull_lvm_by_command
could perhaps use some caching, but it's quite cumbersome to arrange
that, and given that vgs is taking *minutes* to run I think putting
effort into the difference between extremely slow and merely very slow
is not all that exciting, versus fixing the bug at its root.)

I think you should find that a good workaround is to mount /run in the
chroot.  This is normally a good idea anyway if you're doing anything
complicated in the chroot.

Regards,

-- 
Colin Watson                                       [cjwatson at debian.org]



More information about the Pkg-grub-devel mailing list