Re: vivaldi?

["Christian T. Steigies" <cts@debian.org>]

On Sun, Sep 29, 2013 at 11:26:51AM +0200, Ingo J?rgensmann wrote:
> On 2013-09-29 11:09, Christian T. Steigies wrote:
> 
> >E: 05lvm:   Volume group "vg" has insufficient free space (175
> >extents): 512 required.
> 
> Yeah, that's the well-known LVM issue with snapshots on buildds (not
> only on m68k).

Ah, ok, aba mentioned this happens a lot on other arches.
 
> vivaldi:/home/ij# lvs
>   LV                                                   VG   Attr
> LSize  Pool Origin Data%  Move Log Copy%  Convert
>   buildd                                               vg
> owi-aos-- 12.00g
>   sid-m68k-sbuild-223fb7ba-0a0f-4f54-8f3e-5da8bbb00c40 vg
> swi-aos--  2.00g      buildd  17.67
>   sid-m68k-sbuild-a0940ed9-f6ba-409f-a606-b5e2e59d288a vg
> -wi-a----  2.00g
>   sid-m68k-sbuild-c62c1e15-74b6-4c11-9f88-50ef89ce5f45 vg
> -wi-a----  2.00g
>   sid-m68k-sbuild-cb826720-f10c-4aa1-9345-231532fa110c vg
> -wi-a----  2.00g
> 
> >Can somebody fix this?
> 
> There's a tiny script in /root/bin/cleanup-schroots.sh on Arrakis,
> Vivaldi & Elgar to cleanup the mess. But of course it can only be
> run when there's no buildd ongoing. Currently Vivaldi is building.
> 
> The long run fix might be to hope for an upstream fix of the LVM
> problem or to switch to btrfs snapshots. Or to cleanup the cruft in
> pre/post scripts of sbuild.

I think other arches have switched to btrfs or unpacking tarballs. I don't
think I want to work with tarballs on m68k, snapshots are much faster.

I have no idea what these numbers mean. On kullervo I see:
root@kullervo:~>lvs
  LV                                                   VG       Attr LSize Pool Origin     Data%  Move Log Copy%  Convert
  buildd-sid                                           kullervo owi-aos-- 4.00g                                      
  sid-m68k-sbuild-9a14cde0-adc3-4598-9007-f6ada8958e37 kullervo swi-a-s-- 4.00g      buildd-sid  91.59               
  sid-m68k-sbuild-9d323b40-9d1e-440d-a250-279dfad5d006 kullervo swi-a-s-- 4.00g      buildd-sid  84.73               
  sid-m68k-sbuild-a9f9897a-a700-4449-b11c-36ac800b9b23 kullervo -wi-a---- 4.00g                                      
  sid-m68k-sbuild-e0f47be2-9c2a-486d-93e8-8f1612be56f0 kullervo swi-aos-- 4.00g      buildd-sid   1.81               
  user                                                 kullervo -wi-ao--- 4.00g                                     

So the first two snapshops seem to be pretty full... but the partitions are
not mounted, only the e0f4.. one is in use. are these unused partitions the
problem? As soon as there are too many of them, I run out of "extends"?
Since the builds take a few hours on m68k, we do not run into this problem
as fast as ia64 and others. Shouldn't it be possible to remove these
outdated snapshots manually?

Since I booted the unified 3.10 kernel (37 days ago), I have seen only one
segfault on one of the buildds, so this is the most stable kernel in a
really long time! And I have not run out of "extends" yet. Is vivaldi
running 3.10?

Christian