[Libguestfs] [PATCH] New tool: virt-format

Török Edwin edwintorok at gmail.com
Tue Jan 17 18:19:28 UTC 2012


On 01/17/2012 08:11 PM, Richard W.M. Jones wrote:
> On Tue, Jan 17, 2012 at 07:58:15PM +0200, Török Edwin wrote:
>> - does it autodetect when disk is too large for MBR and GPT would be
>>     needed?
>
> No, but you're right that it should do.  I'll make this change.
>
>> - does it align partitions to SSD erase block size? (LVM is
>>     particularly tricky to get right manually)
>
> It aligns to 64K at the moment, for partitions.  For LVM we just use
> the defaults, whatever they are.

I think you need to account for the LVM header size, there was a blog-post from Theodore Tso on this topic:
http://web.archive.org/web/20101130200444/http://thunk.org/tytso/blog/2009/02/20/aligning-filesystems-to-an-ssds-erase-block-size/

>  I don't know if it is possible to
> pass the SSD erase block size up through the virt stack.

I was thinking more about a command-line flag to specify alignment.
My SSD reports 512 bytes as both logical and physical sector size, and thats a physical drive.
Only my HDD reports a 4096 physical sector size (512 logical).

>
> CC'd to Paolo Bonzini: Will virtio-scsi make this easier for us in
> future?
>
>>   - can I specify a backing-file for qemu disks? (qcow, and qed I think)
>
> Yes, in the sense that it just uses the qcow2 disk and doesn't attempt
> to create one, so whatever backing-file etc settings are already on
> the qcow2 disk are not modified in any way.
>
>>   - can --wipe use TRIM on SSDs instead of writing zeroes?
>
> No.  This is a missing qemu feature which I've asked about a few
> times, but until qemu and the rest of the virt stack implements it, we
> won't have it in libguestfs (and conversely when it is implemented,
> we'll most likely get it for free).
>
> Paolo Bonzini: ^^ same question about virtio-scsi ^^
>
>>   - can I setup LVM striping and mirror with it?
>
> No.  In general these are not exposed through libguestfs, although
> they should be, and could be easily enough if someone wrote the patch.
>
> Rich.
>




More information about the Libguestfs mailing list