Stage directories should be deleted in the background, during cron runs

If we do this issue do we really need the PreDestroy and PostDestroy events?

omkar.podey → made their first commit to this issue’s fork.

Needs more info on how exactly this will work does the cron operation just search all instances of DELETE_stage in case of /private/tmp/package_manager_testing_roottest64053336/stage the directory to be searched in is according to the site settings and just starts recursively deleting all things inside of it ? OR should we somehow tell the cron specifically what directory needs to be cleaned up.

I did some simple benchmarking in 📌 In package_manager_test_event_logger also log the time to help benchmarking and debugging Needs work . It looks like deleting the directory does not take take much time compared to the other stages. 4 secs.

So the 4 seconds is better but not a huge improvement. I wonder if the bigger advantage would be removing the PreDestroy and PostDestory events. What would be use cases for using these vs PostApply.

If we remove both of them, I'd be comfortable removing PreDestroyEvent and PostDestroyEvent. That would indeed be simpler, and I can't think of a case for them that can't be done in post-apply.

re #5
I think would just delete everything inside of \Drupal\package_manager\PathLocator::getStagingRoot that starts with DELETE_
Won't the stage directories pile up if cron isn't run for a long time ? do we need to make this optional ?
Cron is expected to be run regularly or your site might not function correctly

though maybe with project browser I could many stage life cycles before cron is run.

Another option we could do is set a set state flag in \Drupal\package_manager\StageBase::destroy and then create StageCleanUp service that subscribes to \Symfony\Component\HttpKernel\KernelEvents::TERMINATE and checks the flag and then deletes any "DELETE_" directories

Looking good. Needs work for MR comments

Looking good. Just a couple small things

looks good.

Looks like 3.0.x needs to be merged in again but otherwise should be good if tests pass again

Automatically closed - issue fixed for 2 weeks with no activity.