We learned a great deal about general satisfaction from the survey performed last October and a bit about usage patterns (usage among a handful of roles). Despite its value, however, survey results don't give us enough information to effectively organize/cull roles or make more informed optimizations (performance varies greatly due to diversity of hardware, platforms, networks).
=== Requirements ===
- Collection and reporting is opt-in (upon setup?)
- Statistics are gather periodically (upon each provision?)
- Data is submitted to a central and query-able store (elasticsearch node?)
- Data is anonymized //before// transmission
- Record:
- Host/guest OS
- CPU arch/model/clock/cores
- Memory available/free/usage (by vbox if possible)
- Non-sensitive MW-Vagrant settings (vagrant_ram, vagrant_cores, nfs_shares, forward_agent, foward_x11)
- Enabled roles
- Number of local roles (would file names violate anonymity?)
- Locally overridden hiera setting names
- Puppet resource types and provisioning times
- More?