Generally puppet-merge works, but I think it would be good to discuss improvements to help take away some of the sharp edges. In particular prevention of conditions that can cause masters to go out of sync with each other.
Creating this paste for async discussion about puppet-merge improvements.
Some ideas...
* Pre-flight checks — Check that destination hosts are up and ready before executing merge commands on any host. Abort if this fails.
* Check that git file ownerships are correct on filesystem. Ensure via puppet, check/fix during puppet-merge, or alert of not.
* Prevent interruption during critical steps.
* Prevent failure on one master from affecting other masters. Have ability to de-pool if failure occurs?
* Alert (maybe even auto-depool by stopping apache?) if puppet repo on master goes out of sync with production for too long.
* An intuitive check/sync feature to check all masters and bring any that are out of sync up to date on demand.
* Improved help text. Perhaps also change default action of script to print help and require an option to run the merge process.