To add to the current advice, I would enable debug logging and monitor Foreman’s production.log.
This old post has some nice tips from lzap for looking into memory issues: Puma cluster worker using huge amounts of memory - #3 by jjvcuyler
In the log, with debug logging on, seeing what controllers are call frequently could give some direction.
Like:
2025-12-10T13:52:46 [I|app|55c3c877] Processing by Katello::Api::V2::HostPackagesController#containerfile_install_command as JSON
...
2025-12-10T13:52:46 [I|app|55c3c877] Completed 404 Not Found in 151ms (Views: 0.2ms | ActiveRecord: 23.6ms | Allocations: 50900)
Are there log records with large timing results? For example:
2022-01-25T05:14:55 [I|app|45ccd5b1] Completed 201 Created in 4909255ms (Views: 102.3ms | ActiveRecord: 352203.4ms | Allocations: 553340631)
These can help point to inefficient actions.