[Ovirt-devel] [PATCH] Added a migration.

Hugh O. Brock hbrock at redhat.com
Mon May 19 15:19:51 UTC 2008


On Mon, May 19, 2008 at 10:08:12AM -0400, Scott Seago wrote:
> Chris Lalancette wrote:
>> Scott Seago wrote:
>>   
>>> Long term, once we've got to support upgrades, that's the way we've got 
>>> to go. The question is when do we switch over to that model. Is now as  
>>> good a time as any, or do we keep editing the migrations until we've at 
>>> least reached a stable functional demo?
>>>
>>> Any thoughts?
>>>     
>>
>> I'm of two minds about it.  On the one hand, I don't think we necessarily want
>> to have the additional development burden of doing migrations while things are
>> still in flux.  On the other hand, it sure would make updates between release
>> 0.x and 0.x+1 a hell of a lot smoother, not to mention updating from release 0.x
>> to git head (basically just need to run rake db:migrate).  How much extra work
>> is it to add migration code when the database is changed?
>>
>> Chris Lalancette
>>   
> It's not much more work than changing the existing migration when you don't 
> care about data preservation. For the latter, it could be potentially 
> complex. In the example that triggered my question -- our current way of 
> doing this would just be to replace 'user' with 'uid' in one migration 
> file. This approach required a new migration, but a rather simple one -- 
> basically a single rename_column operation.
>
> Scott

I think we should avoid encumbering ourselves with migrations until the
summit release. After that we should absolutely start doing them.

Take care,
--H




More information about the ovirt-devel mailing list