[pve-devel] Dedicated Migration Network #1177 v2

[Thomas Lamprecht]

On 11/18/2016 06:28 PM, Phil Kauffman wrote:
> On 11/17/2016 04:48 AM, Thomas Lamprecht wrote:
>> Does this trigger always?
> Yes.
>
>
>> This seems not directly related to the migration network changes,
>> as the API call _can_ already connect to the other node over the
>> specified network.
>> Else it would not output like:
>>
>>  > Nov 17 11:12:40 # /usr/bin/ssh -o 'BatchMode=yes'
>> root at fd0f:38:500:15::3 /bin/true
>>  > Nov 17 11:12:40 Host key verification failed.
>>
>> And if he can connect he does exactly the same things as on the other
>> network.
>>
>> Can you try migration type 'insecure' also?
> Yes. I have tried both secure and insecure.
>
>
>> Seems like a problem with your other network? Not quite sure, that seems
>> strange to me.
> This is possible, I am continuing to debug on my end with some test 
> machines.
>

Let me hear if you got some new results.
A few people tested this here successfully, so at least basic 
functionality should work.

>
>>> ### Migration succeeds, but parse error on /etc/pve/datacenter.cfg
>>>
>>> root at bravo-vmnode2:~# cat /etc/pve/datacenter.cfg
>>> migration: network=192.168.2.0/24
>>
>> You're using a wrong format here, if the migration property gets set the
>> type needs always to be set too.
> I know. The man page indicates otherwise.
>
> migration: [type=]<secure|insecure> [,network=<CIDR>]
>            For cluster wide migration settings.
> I also states the type=secure is the default, also indicating that 
> 'type' is optional.
>
>

Yes, it's a bit misleading. The indicator on [type] just means that you 
may omit type, not the whole param.
I marked the type parameter as default_key so that
migration: insecure
is the same as
migration: type=secure

this has the disadvantage that you cannot omit it. Should not do that 
anymore, my bad :)

cheers,
Thomas