Current TOP wakes itself once a process is about to be completed. This happens by sending itself a message as soon as the process is completed (based on Schedule.forecast/1).
This works fine but has a big problem:
VM crashes / reboot (every Helix deployment reboots the VM).
While it's possible to handle most cases gracefully, it doesn't hurt to persist the process on disk. Every time the VM boots, it should look for all processes which are/were being tracked. If any of these has expired their estimated time (i.e. they are completed), then it should call TOPAction.complete/1. For all other processes which are still running, it should use the standard TOPBringMeToLifeEvent.
The implementation of this persistence layer was removed from #322 and delegated to another PR.
Current TOP wakes itself once a process is about to be completed. This happens by sending itself a message as soon as the process is completed (based on
Schedule.forecast/1).This works fine but has a big problem:
While it's possible to handle most cases gracefully, it doesn't hurt to persist the process on disk. Every time the VM boots, it should look for all processes which are/were being tracked. If any of these has expired their estimated time (i.e. they are completed), then it should call
TOPAction.complete/1. For all other processes which are still running, it should use the standardTOPBringMeToLifeEvent.The implementation of this persistence layer was removed from #322 and delegated to another PR.