On 08/24/2016 12:45 PM, Alan Pevec wrote:
>>> If we can go back just one commit, to "commit 2 to openstack-cinder", and it is passes CI, we could still have a consistent repo promoted, just not the very last one at the time of running the CI job.
>>> Actually, this makes sense to me. By doing this, we could promote the last known good+consistent repo, while we fix the latest issue.
>> Yes that was the intent. Let's see how we could implement this.
> Hold on, haven't we previously discussed to switch to staged promotion
> i.e. puppet CI promotes RDO Trunk consistent first, then tripleo CI
> tries to promote that and finally RDO CI uses what tripleo CI
> promoted? Now we have all three starting from different random latest
> "consistent" (at the time corresponding promotion runs) and diverging.
> But in any case, first step is to have a database of "consistent"
> hashes recorded somewhere, it would be just input to the puppet CI
> promotion proposal script, right?
The staged promotion is a different story that we need to continue. For
me that's a parallel effort we need to pursue. But even with a staged
promotion, we could still want to try on a failure a previous consistent
repo that we could promote.