The problem

When you create and destroy nodes in your puppet infrastructure you build up a lot of deactivated nodes in your PuppetDB database that serve no real purpose but do slow down the performance of the database and cause it to grow.

Suggested Solution

node-purge-ttl should be enabled by default and set equal to or higher than report-ttl. It should generally be set to an interval longer than report-ttl to avoid needing the node purge process to also delete reports.

Considerations

It could be that we need to add a migration step that will delete old nodes during an upgrade so that post-upgrade the service isn't churning through deleting nodes but not processing catalogs or facts.

Upon further thought, it would probably be better to implement PDB-3173 and have the default be that node-purge-ttl will only delete up to 100 nodes whenever it runs. Then when this setting is enabled it shouldn't suddenly cause too much churn if you have 1000s of deactivated nodes.