[Libvirt-cim] [PATCH 0/3] cimtest follow patch

[John Ferlan]

On 04/10/2013 10:51 PM, Wenchao Xia wrote:
>> On 04/10/2013 04:26 AM, Wenchao Xia wrote:
>>> 于 2013-4-9 23:45, John Ferlan 写道:
>>>> On 04/09/2013 10:40 AM, John Ferlan wrote:
>>>>> On 04/08/2013 06:16 AM, Wenchao Xia wrote:
>> I'm *painfully* aware of #2 & #3!!!  The "default" libvirt location is
>> also /var/lib/libvirt/images and cimtest kept failing strangely until
>   Me too, maybe we can skip this now and improve it in the future.
> 

Right - I'm not going to solve the problem - just have it noted.


>   How do you installed libvirt-cim? I guess "make install" still
> have problem, could u try remove old libvirt-cim rpm and then "make rpm"
> in libvirt-cim source code and then rpm -ivh?
>   Also have you installed pywbem rpm? softlink seems work innormal since
> pywbem have dependency library need to be installed.

What's interesting about the RH64 system is I used yum/rpm install
only...  The rest is a bit lengthy as I've added to it during the day,
but perhaps it's good to know/record the steps/process used.

I yum localinstall'd libvirt-cim... pywbem was installed... I erased and
reinstalled libvirt-cim today (sometimes that helps) and yet still the
MigrationService was still missing, although MigrationCapabilities
miraculously showed up.

A 'wbemcli ecn http://root:password@localhost/root/virt | grep
Migration' did not find a CIM_VirtualSystemMigrationService.

After digging through the various scripts I believe that means the
{KVM|LXC|XEN} versions of the mof won't be installed/loaded

Why it's not there - I have no clue. Per the libvirt.org/CIM/schema
webpage, I installed the v216 experimental schema on my RH64 box. That
schema doesn't have that class/mof; however, on my F18 box it seems I
have a later schema installed (v2.33) which does have it. Not sure how I
installed that... According to everything I found online the migration
schema was added in v2.17.

After spending some time thinking about and investigating the
differences between the two systems - I tried using the 'make
{preinstall|install|postinstall}' options and had some interesting
results (with a restart of tog-pegasus between each make option).
Interesting as in things now work (except for one small bug in
'enum_volumes' where 'None' was returned for an empty storage pool
causing an exception - I have a fix for it). Of most interest is somehow
CIM_VirtualSystemMigrationService is now present. What I noted as part
of that installation which went to /usr/lib64/share/libvirt-cim is that
there is a 2.21 schema zip file in that directory which gets inflated
and installed as part of the 'make' processing (by one of the two
scripts).  Since 2.21 > 2.17 that migration mof is there, so now it's on
that system. Why this doesn't happen as part of RPM install - I'm not sure.

So to summarize -
1. Now I have no test failures on my RH64 system other than the already
failing Indications tests.

2. I believe there's some disconnect between what happens via the rpm
install and what happens during the 'make' options, but I don't know
where to look and right now I really don't have the cycles to
investigate. I'm guessing that somewhere along the line 2.21 was made
the default, but the web pages didn't get updated and something in the
RPM install process didn't quite work right, but it didn't matter or
wasn't noticed because perhaps no one went through the pain of a clean
installation environment while strictly following the web pages.

John