[RFE] Caching could be after power-on

[Julia Kreger]

Presently, when a deployment (via the iscsi deployment interface) is initiated where the local image cache is used, the deployment waits upon the local cache to have a copy of the image, before proceeding. For discussion purposes, say that image download takes 7 minutes from start to finish.

At the same time, we also wait to power-on the node until the cache is populated. The downside with this is with hardware that may take, say 7 minutes, just to POST.

At which point, IPA may begin to boot but not be ready yet, and we've already spent 14 minutes where there could have possibly have been some sort of concurrency. For a deployment that takes a minimum of 21 minutes for a node to reach active, shaving 7 minutes off is a major improvement.

Naturally, there may need to be tuning in IPA with how it begins heartbeating, in that a "deployment pending cache" state for the image might cause IPA to heartbeat far faster than normally.

[Ruby Loo (rloo)]

I'm not sure that this needs a spec -- depends on the approach taken to do this and how complicated it may be. So either add a spec or update the description with more details as to what this concurrency entails.

/me wonders whether there be config options involved... (in which case, yes, a spec...)

I agree that shaving 7 minutes is worth it :)