[Libguestfs] Proposal to remove two virt-v2v command line options

[Roman Kagan]

On Mon, Apr 25, 2016 at 07:54:36PM +0100, Richard W.M. Jones wrote:
> Virt-v2v aims to convert guests without any per-guest intervention/
> configuration/fiddling.
> 
> Unfortunately there are some command line options of virt-v2v which
> don't meet this ideal.  Although we don't normally change the command
> line of our tools, there are two options I'd like to remove (by
> turning them into warnings & no-ops).
> 
> Please follow up with comments if this is going to be a problem.
> 
> 
> Option:
>    --no-trim all
>    --no-trim mp[,mp...]
> 
> Description in manual:
> 
>        By default virt-v2v runs fstrim(8) to reduce the amount of data
>        that needs to be copied.  This is known to break some buggy
>        bootloaders causing boot failures after conversion (see for example
>        https://bugzilla.redhat.com/show_bug.cgi?id=1141145#c27).
> 
>        You can use --no-trim all to disable all trimming.  Note this will
>        greatly increase the amount of data that has to be copied and can
>        make virt-v2v run much more slowly.
> 
>        You can also disable trimming on selected filesystems only
>        (specified by a comma-separated list of their mount point(s) in the
>        guest).  Typically you would use --no-trim /boot to work around the
>        grub bug mentioned above.
> 
>        You can also disable trimming on partitions using the libguestfs
>        naming scheme for devices, eg: --no-trim /dev/sdb2 means do not
>        trim the second partition on the second block device.  Use
>        virt-filesystems(1) to list filesystem names in a guest.
> 
> Discussion:
> 
> As noted in the description, we found one RHEL 5 guest which had a
> buggy bootloader.  We've never been able to repeat that or find a
> similarly broken guest.
> 
> When I added this option, I was suspicious that fstrim would cause
> other breakage in guests.  With the popularity and widespread use of
> virt-sparsify, I think that fear was unjustified.
> 
> Supporting the --no-trim option is also a pain, and it's difficult
> even for users to understand what it does, so I propose we remove it.
> Note this means that all guests would be fstrimmed.
> 
> 
> Option:
>    --vmtype desktop
>    --vmtype server
> 
> Description in manual:
> 
>        For the -o rhev or -o vdsm targets only, specify the type of guest.
>        You can set this to "desktop" or "server".  If the option is not
>        given, then a suitable default is chosen based on the detected
>        guest operating system.
> 
> Discussion:
> 
> This option was inherited from old virt-v2v and it sets a single flag
> in the OVF file when targetting oVirt/RHEV.  As far as I can tell this
> flag affects two things in oVirt engine:
> 
> (1) Whether sound emulation is enabled for a guest ("desktop" = yes,
> "server" = no).
> 
> (2) allowConsoleReconnect which disables something called "strict user
> checking" when connecting to a VM console ("server" = disabled,
> "desktop" = not disabled).
> 
> It is highly unlikely that a virt-v2v user understands this setting
> (even I didn't understand it before now).  Note that we choose a
> default from inspection data, mapping guest server-like operating
> systems to "server" and the rest to "desktop".
> 
> So I propose we get rid of this option, but leave the code which
> performs the default mapping from inspection data.

FWIW we at Virtuozzo neither use nor plan to use these options so we're
not affected by their removal.

Thanks,
Roman.