After talking with @csteipp, it's clear that the road to full productionization of the Program Dashboard Rails application—deployment to the production cluster, after deployment to labs by the end of this month—is going to entail a substantial amount of review, coordination, and some level of long-term ops commitment/support. Let's get this all on the table before the month is up, so we can communicate the realities to our stakeholders (and other jargon).
- Identify immediate options for and limitations to running more reliably in labs (database consistency, replication, failover, varnish cache, etc.)
- Identify security review requirements, commitments so we can scope a rough long-term productionization timeline.
- How can we manage Rails application dependencies to allow for security review while maintaining a level of flexibility in deployment?
- Bundler packaging of gems?
- Precompilation of native extensions?
- What will definitely require Debian packaging, etc.?
- Are there options for some level of isolation in production that will mitigate security concerns? (Ganeti?)
- What level of Ops commitment would we need to run a Rails application in production long-term? Can we get that commitment elsewhere?