Greetings,
I periodically rebuild a personal Fedora remix to have all of the updates... and am not
working with Fedora 16. In the x86_64 version of my remix I include all of the KVM
virtualization stuff. It seems that a recent set of updates has caused zfs-fuse to become
a dependency where it wasn't previously. That's fine... but I noticed that my
remix would fail to build near the tail end of the process. Since the initial failure
there have been a few more waves of updates and I've still been unable to complete a
rebuild.
I finally figured out how to work around the issue. When zfs-fuse gets installed inside
of the build root it appears to start up on the build host. By killing the top level
zfs-fuse process manually with kill and making sure it is no longer running, I was able to
finish the build process. From the host node context zfs-fuse was not enabled and the
systemctl stop zfs-fuse would say it wasn't running... I'm guessing because it was
installed and started up by the livecd-creator build process rather than on the host node
proper. It appears that if zfs-fuse is allowed to stay running the build process will
fail but by manually killing it ('pstree -nup | less' search for parent zfs-fuze
process and kill it) it will complete. I haven't really looked for a way to automate
the kill and just do it manually while the remix is building once I notice that the
zfs-fuse service is running.
I haven't looked for nor filed a bug because the path of blame is unclear to me... and
I don't know that it affects many Fedora users in the wild other than us respin /
remix builders... but I thought I'd pass it along for anyone else who might care.
The only other time I've run into an issue with a particular package breaking the
build process was with google-chrome browser. Over a year ago having google-chrome
browser listed in my remix package list, when google-chrome gets installed it does a
service restart (I don't recall which one, atd maybe?) and coincidentally the remix
build process would fail near the tail end. I do not know if that is the case now because
at that point I took google-chrome out of my package list and I have yet to add it back.
It certainly is not a Fedora provided package so I wouldn't expect Fedora to
troubleshoot such an issue.
Any any event, from those two incidents it has lead me to believe that if a package
post-install script does something to the state of services, that the remix build process
will probably fail. I'm not sure who cares or if they'd like further
investigation in an effort to validate the problem and work towards a solution or not...
but anyway... feel free to comment if desired.
TYL,
--
Scott Dowdle
704 Church Street
Belgrade, MT 59714
(406)388-0827 [home]
(406)994-3931 [work]
Show replies by date