On Sat, May 04, 2019 at 06:59:22PM +0100, Colin Watson wrote: > On Fri, Jan 18, 2019 at 03:16:26AM +0100, Andreas Bombe wrote: > > This has worked for all these years until the update, which rejects the > > VG with system ID without advance warning during package upgrade. Systems > > may become unbootable after upgrade if these VGs contain filesystems > > that are mounted in /etc/fstab. > [...] > > The only way to get access was to set global/system_id_source = > > "lvmlocal" and set local/system_id to the affected VG's system ID. > > I also ran into this on upgrading from stretch to buster. Fortunately I > was keeping enough of an eye on console output from the upgrade that I > was able to find this bug and I worked around it as you describe here.
In case it's useful information to anyone, since the system ID appears to contain a seconds-since-epoch value, I can deduce that I created the VG on 2004-05-18, which makes sense given the sorts of things I was doing at the time. I would have been running either woody or the current state of sarge in development at the time. I added the VG to an existing system rather than installing a new one, so the installer wasn't involved. I suppose it's technically possible that I set the system ID myself, but it's in the form <hostname><creation-seconds-since-epoch> which I don't *think* is a format I would have used as my first pick (I'm sure I'd at least have included some kind of separator); it seems much more likely that it was the default for vgcreate at the time, although I haven't yet found anything relevant in any changelogs. I have a VG elsewhere created in early 2013 that doesn't have a system ID, and friends report similar things back as far as squeeze, so I guess that narrows it down a bit. -- Colin Watson [[email protected]]
