Il 24/07/2013 12:31, Daniel P. Berrange ha scritto:
On Wed, Jul 24, 2013 at 12:25:17PM +0200, Paolo Bonzini wrote:
> Il 24/07/2013 11:54, Daniel P. Berrange ha scritto:
>>> but anyway, we provide good enough documentations, so if one wants to
>>> use "direct" mode, he can still change the xml to use it, any big
problems
>>> here? though defaulting to "direct" is more convenient for these
users.
>>
>> Defaulting to "direct" is *NOT* more convenient for all users, only
>> those who happen to have a new enough QEMU, from a vendor who actually
>> decided to compile in the iSCSI support. If they're lacking such a QEMU
>> then the "direct" mode will prevent the default config from even
starting.
>> This can never be an acceptable default choice. mode="host", while it
huas
>> some limitations in features, is the only option that is capable of working
>> with any QEMU
>
> s/limitations in features/possibility of wreaking havoc to clusters/
This isn't a new issue. People have been using KVM with iSCSI in host
mode for as long as KVM has existed and the sky hasn't fallen in
clusters. As long as it is a documented limitation, that admins can
address with a config setting if desired, I don't see that this is
suddenly something we must fix at the cost of breaking the default
config for QEMU's without iSCSI support.
Understood, that's why I suggested not having a default config at all
for type='lun' devices.
Also, can libvirt at least detect using a volume on a pool that wasn't
started, and fail unless mode='direct'?
Paolo