similar to: [FOR REVIEW ONLY] ESX work in progress

With this patchset I have successfully[1] imported a RHEL 5 guest directly from ESX with the following command line: virt-v2v -ic 'esx://yellow.marston/?no_verify=1' -op transfer RHEL5-64 Login details are stored in ~/.netrc Note that this is the only guest I've tested against. I haven't for example, checked that I haven't broken Xen imports. Matt [1] With the exception of

These patches combine HVSource and HVTarget into a single Converter. This should make it more obvious where to hack without losing any practical flexibility. GuestOS remains separate. GuestOS is now a misnomer, because it's really only a Linux distro abstraction. It will be useless for Windows, for example. Functions which you'd expect to be different on a non-RH distro should live in

These 2 patches are mostly code motion. They were prompted by an apparent augeas error in BZ 613967 which didn't display useful error message. The error seems to happen in Converter::Linux. GuestOS::RedHat had a handy function which displayed verbose augeas error messages. This function moves into the new module where it can be used by both modules. The second patch is an consequential tidy

We now always copy a guest during conversion, meaning this tool is no longer required. --- Build.PL | 2 +- MANIFEST | 2 - lib/Sys/VirtV2V.pm | 9 +- lib/Sys/VirtV2V/Connection.pm | 1 - po/POTFILES.in | 1 - snapshot/run-snapshot-locally | 43 -- snapshot/v2v-snapshot.pl | 931

I noticed that virt-v2v, which is written exclusively in perl, failed to generate virt-v2v.pot. After much head scratching I also noticed that libguestfs.pot didn't include any messages from perl sources. Some reading of libintl-perl shows that a somewhat more complicated xgettext command line is required, as it doesn't understand Locale::TextDomain syntax by default. After a little more

Previously, only the LibVirtXML connection translated network and bridge names in imported metadata. This change moves this functionality in Converter, making it available to LibVirt connections as well. At the same time, the format of the config file is switched to XML. The primary driver for this is that the allowable syntax of a foreign network/bridge name is not known. Rather than create a

If a user accidentally passed a valid XML document to the LibvirtXML which wasn't actually a libvirt domain XML document, they would end up getting obtuse errors from a libguestfs library. This change does 2 sanity checks: 1. After parsing a the XML document, check /domain/name exists. 2. Check that the domain defines some storage The second check is generally useful, as we can't

Previously, only the LibVirtXML connection translated network and bridge names in imported metadata. This change moves this functionality in Converter, making it available to LibVirt connections as well. At the same time, the format of the config file is switched to XML. The primary driver for this is that the allowable syntax of a foreign network/bridge name is not known. Rather than create a

Move all target-specific functionality into its own module in preparation for output to RHEV. --- MANIFEST | 1 + lib/Sys/VirtV2V/Connection.pm | 46 ++--- lib/Sys/VirtV2V/Converter.pm | 138 +------------ lib/Sys/VirtV2V/Target/LibVirt.pm | 419 +++++++++++++++++++++++++++++++++++++ lib/Sys/VirtV2V/Transfer/ESX.pm | 91 +++------ po/POTFILES.in

Also changes command line parsing to require a pool to be specified when using libvirt output, meaning storage will always be copied. --- MANIFEST | 1 + lib/Sys/VirtV2V/Connection.pm | 74 +++++++++--------- lib/Sys/VirtV2V/Connection/LibVirt.pm | 18 +++-- lib/Sys/VirtV2V/Connection/LibVirtXML.pm | 11 ++- lib/Sys/VirtV2V/Target/LibVirt.pm

Cleanup was not happening properly if a migration to RHEV was killed prematurely with a Ctrl-C. Firstly, the SIGINT and SIGQUIT handlers were not being registered early enough in virt-v2v.pl. Secondly, if Ctrl-C killed the guestfs qemu process first it would deliver a SIGPIPE to v2v, which caused an unclean shutdown without cleanup. Fixes RHBZ#596015 --- v2v/virt-v2v.pl | 17 ++++++++++++++---

Exit with an error if we would overwrite an existing libvirt domain. Fixes RHBZ#617110 --- lib/Sys/VirtV2V/Connection/LibVirt.pm | 4 ++++ lib/Sys/VirtV2V/Target/LibVirt.pm | 31 ++++++++++++++++++++++++++++++- lib/Sys/VirtV2V/Target/RHEV.pm | 11 +++++++++++ 3 files changed, 45 insertions(+), 1 deletions(-) diff --git a/lib/Sys/VirtV2V/Connection/LibVirt.pm

rmtree can die instead of returning failure under some circumstances. We don't want this to stop cleanup. --- lib/Sys/VirtV2V/Target/RHEV.pm | 15 ++++++++++++--- 1 files changed, 12 insertions(+), 3 deletions(-) diff --git a/lib/Sys/VirtV2V/Target/RHEV.pm b/lib/Sys/VirtV2V/Target/RHEV.pm index 295c19d..c9add0c 100644 --- a/lib/Sys/VirtV2V/Target/RHEV.pm +++

We were seeing 100% failure rates transferring 10G disk images from ESX on a particular setup. We also weren't spotting the transfer failure, and dying with a strange error from libguestfs. These 2 patches fix the error check which should have made it obvious what was failing, and the underlying error.

Although I can't reproduce, I seem to have had an incomplete transfer which wasn't picked up by the existing error handling. This change adds some more sanity checks: * Check for error on close() (should have been doing this anyway) * Check that the bytes received == reported content length --- lib/Sys/VirtV2V/Transfer/ESX.pm | 14 +++++++++++++- 1 files changed, 13 insertions(+), 1

Unfortunately libvirt will not always return volume information for a guest's volume. Specifically, it will not return volume information if the volume is not in a storage pool. Rather than displaying a generic libvirt error in this case, advise the user how they can create an appropriate storage pool. Fixes RHBZ#672498 --- lib/Sys/VirtV2V/Connection/LibVirtSource.pm | 27

This patch series includes a repost of Milan's unmodified RHN support patch because I haven't pushed it yet. On top of that patch, it includes the capabilities patch in as many bits as I could make it into. The big one is 7/8. I've tested all of the following guests both with and without RHN registration: Xen RHEL 54 PV Xen RHEL 51 PV Xen RHEL 48 PV ESX RHEL 54 FV

We were assuming that guests wouldn't use multiple interfaces, or if they did, drive letters wouldn't clash between the interfaces. An ESX guest with a CDROM device breaks this because hard disks will be presented a SCSI, starting at sda, and CDROM devices are presented as IDE, starting at hda. This is in contrast to QEMU, which starts at hdc by default. Firstly, this change causes disks

If a volume is not an exact multiple of 512 bytes, qemu-img will report its size rounded down to a 512 byte boundary. However, when copying, the file is still read until EOF, which will return more data than was expected. This change prevents that causing a failure in itself. The situation is still not resolved, however, as there are still situations where this will cause a failure. For example,

This patch refactors the data transfer code with several goals: * Have a common read(source)/write(target) loop so that common processing can happen in the middle of it, e.g. format change/progress bar * Provide volume metadata to transfers to allow smarter reading/writing, e.g. of sparse files * Simplify the data transfer code The patch *isn't* NFC because there are some minor behaviour