[libvirt] [PATCHv2 1/7] interface: new public API for networkconfig change transactions

[Laine Stump]

On 05/19/2011 06:15 PM, Christian Benvenuti (benve) wrote:
>> -----Original Message-----
>> From: libvir-list-bounces at redhat.com [mailto:libvir-list-
>> bounces at redhat.com] On Behalf Of Laine Stump
>> Sent: Thursday, May 19, 2011 1:51 PM
>> To: libvir-list at redhat.com
>> Subject: [libvirt] [PATCHv2 1/7] interface: new public API for
>> networkconfig change transactions
>>
>> From: Michal Privoznik<mprivozn at redhat.com>
>>
>> This is the API agreed on in:
>>
>>    https://www.redhat.com/archives/libvir-list/2011-May/msg00026.html
>>
>> (with a slight name change to use "...begin" rather than
>> "...start"). This implements transactional changes to the host network
>> config. When a transaction is begun with ncf_change_begin(), all other
>> netcf APIs will continue to work as they always have, but a snapshot
>> of the existing config will be taken, allowing reversion (rollback,
>> using ncf_change_rollback()) to the exact state of config at the time
>> ncf_change_begin() was called. Alternately, if it's determined that
>> the new changes are acceptable, ncf_change_commit() can be called,
>> which will eliminate the snapshot and make the changes permanent.
>>
>> As a failsafe measure, if neither ncf_change_commit() or
>> ncf_change_rollback() is called by the next time the system reboots,
>> the netcf-transaction initscript will be automatically called to
>> rollback the changes.
> Why do you think the default rollback (post reboot, for non-committed
> transactions) is a good default?

It is intended as a failsafe device when the host becomes unstable, and 
this is the safest default behavior. If someone has gone to the trouble 
of beginning a transaction, that implies that they are nervous enough 
about the outcome of the changes that they think they might need to be 
reverted. And it also implies that they will be monitoring the changes 
and have ample opportunity to commit the transaction prior to the system 
rebooting.

> When/where is that automatic rollback enforced?

It's done at boot time in an initscript that will be installed by netcf. 
See:

https://fedorahosted.org/pipermail/netcf-devel/2011-May/000562.html

> (I could not find it in the patches).

Almost all the functionality lives in the netcf library. the libvirt API 
just provides a simple pass-through to that functionality.

>   Does it trigger a log message too?

Ah - that's a good point! the initscript should log a message indicating 
that it is restoring something rather than just silently doing it. I'll 
make sure to squash that in before pushing the initscript to netcf.

> Do you think it may make sense to have that default behavior controlled
> by a (global, ie, not per-interface) parameter?

So far you're the first person to ask for that. There's nothing in the 
API that would prevent us from adding it in, but my current inclination 
is to get this pushed with the current functionality so people can start 
using it.

> I am not suggesting the introduction of such parameter (but I would not
> be against it), but I would suggest a 4th virsh command to list at
> least the interfaces with a pending (ie, not committed yet) transaction.

Another interesting idea, but something that, as we don't currently have 
a specific need for it, can be done at a later time.