[virt-tools-list] [PATCH] fix: do cleanupLocation only at last time for cached fetcher

Cole Robinson crobinso at redhat.com
Wed Sep 9 12:07:00 UTC 2015

On 09/09/2015 01:56 AM, Chunyan Liu wrote:
> Regression introduced after commit d8d6af5. Failed to install guests
> through ftp network source, see below:
> /usr/bin/virt-install   --name sled-12-fcs-64-pv-nnm-nete5250fdfecc94f922b5a21a
>  -p --location ftp://xen100.virt.lab.novell.com/install/sled12/x86_64
>  --extra-args "autoyast=
> sled-12-fcs-64-pv-nnm-nete5250fdfecc94f922b5a21a"   --disk path=/var/lib/
> libvirt/images/sled-12-fcs-64-pv-nnm-nete5250fdfecc94f922b5a21a.qcow2,size=10,
> format=qcow2   --network=bridge=br0   --ram=1024   --vcpu=2   --vnc
>  --serial pty   --noautoconsole   --video cirrus
> ERROR    'NoneType' object has no attribute 'sendall'
> Domain installation does not appear to have been successful.
> If it was, you can restart your domain by running:
>   virsh --connect xen:/// start sled-12-fcs-64-pv-nnm-nete5250fdfecc94f922b5a21a
> otherwise, please restart your installation.
> Root cause is: now we use cached fetcher, but still do cleanupLocation
> each time, it will cause problem when using the fetcher next time.
> We should only do cleanupLocation at last time.
> Signed-off-by: Chunyan Liu <cyliu at suse.com>
> ---
>  virtinst/distroinstaller.py | 7 +++++--
>  1 file changed, 5 insertions(+), 2 deletions(-)

I can't reproduce the issue, and in fact there isn't any call to 'sendall' in
upstream virt-manager code so not sure what that error is about. Can you
attach the full virt-install --debug output?

>From what I can see, the code is fine: every cleanupLocation call is preceded
by a setupLocation call, so things should work as expected. Your patch against
the current code means that cleanupLocation may never be called depending on
when things error, which could leave stranded mount points depending on the
passed media.

- Cole

More information about the virt-tools-list mailing list