GitLab

Update ChangeLog and version files for snf-image 0.15

Update ChangeLog and version files

Replace all references of `environmental variable' with `environment
variable'

Handle better possible backend options and require necessary
ones by printing appropriate error messages.

Update Changelog and version files.

Synnefo version 0.15.1 has introduced a selectable storage backend
for Pithos, resulting in necessary changes to the ModularBackend class.
In order to keep backward compatibility with older Synnefo versions,
we check the Pithos backend version and instantiate the ModularBackend
class appropriately.

Add multiple version files to automatically update the docs version too.
We need devflow 0.13rc1 or later for this to work

Add Chrysostomos Nanakos to the list of snf-image authors

* Add (k)ubuntu 14.04 and remove 13.10
* Add Oracle Linux
* Remove Debian Squeeze
* Add OpenBSD 5.5 and remove 5.4

In order to extend a partition we need to remove and recreate it. If the
partition was misaligned then the image is to blame. snf-image is not
allowed to change the starting sector of the partition since this will
destroy the file system it hosts. In order to make sure this does never
happen we need to set the sector alignment multiple to 1.

Set default values for Pithos storage backend type, RADOS conf file,
RADOS map and block pools.

Include new Pithos variables for selectable storage backend to Ganeti
create script.

Introduce new Pithos variables for selectable storage backend.

When copying the image data to the VM's hard disk, use iflag=fullblock
to accumulate full blocks of input. Without this flag, dd will write
the read data whenever read() syscall returns, without waiting to gather
enough data to create a bs-sized block. This may lead to unoptimized
performance.

Pass -s flag to kpartx to fix a potential race condition when deploying
(ext|ntfs)dump images. Without -s, kpartx will return right after
signaling the device mapper to create the partition block devices. This
does not guarantee that the partitions will have been created when
snf-image will try to use them.

Add support for resizing UFS file systems of NetBSD OSes

When enlarging OpenBSD partitions, round them up to multiples of 16512
which is the default fs blocksize. This is what OpenBSD's disklabel
does when creating partitions.

The mapping we used would fail in some cases. If a BSD partition
has the same boundaries with an MBR one, the kernel will not map it
onto a different block device. As a result, partition 'a' is sometimes
mapped onto /dev/sda1 and sometimes onto /dev/sda5.

Add an input option in disklabel.py to print the partition mapping the
Linux kernel would do on the disklabel.

 * Implement new NetBSDDisklabel and FreeBSDDisklabel classes that
   subclass BSDDisklabel class.
 * Add methods for enlarging the disk and the last partition in
   NetBSDDisklabel

This method converts an LBA address to CHS. Different BSD systems do
this in different ways:

 * NetBSD uses LBA-Assisted translation but OpenBSD does not.
 * For addresses over 8GB where CHS will overflow, OpenBSD will assign
   a fixed value on all 3 CHS fields when NetBSD will do this only for
   the cylinders field.

Make PartitionTableBase and DisklabelBase classes abstract by using the
ABCMeta metaclass of the abc module

 * Add missing method comments
 * Inline all MBR fields comments
 * Print partition type in hex

Add basic support for traditional BSD disklabels