Timeouts when deleting a post

[pkfrank]

Describe the bug
As a user, I want to be able to Delete my posts. Currently, it occasionally times out.

To Reproduce
This seems to happen more often with articles that have a lot of reactions/comments. So, to reproduce, choose an article you're willing to delete with many reactions/comments and press Delete from your dashboard.

Expected behavior
The article is deleted without timeouts.

[rhymes]

I believe this is due to this line https://github.com/thepracticaldev/dev.to/blob/master/app/controllers/articles_controller.rb#L128

destroy! calls in a cascade all destroy methods of all related objects that can be deleted. If it's a lot of content it will take some time. Given that deletion is done in process (which means the user has to wait for everything to be deleted during their HTTP request) and given that Heroku times out at 30 seconds, this is a likely reason why big trees of content (the article, plus many comments and many reactions and notifications) can go past the limit.

In addition there's this before_destroy_actions that executes before all of what I mentioned before, I guess this executes synchronously as well.

Can I be assigned to this?

[maestromac]

Yep @rhymes ! give it a whirl.

[tylermcginnis]

Getting this as well.

[Zhao-Andy]

This is still an issue, so raising it up again and adding a few details:

We could make the destroy action in the controller run asynchronously, but I don't think that's ideal. A better solution might be to write a service object whose sole responsibility is to handle destroying articles (and all the dependents).

A few -- but not exhaustive -- list of things we'll need to consider:

1. Properly purging all caches of the article, comments on the article, and possibly any cached routes (/reactions?article_id=123 for example)
2. Removing the article and comments from Algolia's index
3. Handling any errors that might happen

[rhymes]

Hi @Zhao-Andy I haven't actually looked into it after all, I'm sorry. I don't think it's an easy PR for me because it requires familiarity with the caching system or Algolia and I don't have that.

Maybe you can give a read to the analysis of the bug I wrote back then on dev.to.

There are a services and aspects to consider: caches, fastly, algolia, stream.io and probably others

I think the best solution is to do a two phase removal:

1. the user needs the article to be gone immediately
2. everything else (caches, algolia, fastly, stream.io maybe as well, dependent objects in the DB) can be removed out of process in the service object that tracks everything and cleans up

[stale]

Thanks for contributing to this issue. As it has been 90 days since the last activity, we are automatically closing the issue in 7 days. This is often because the request was already solved in some way and it just wasn't updated or it's no longer applicable. If this issue still requires attention, please respond with a comment. Happy Coding!

[rhymes]

still an issue

[mstruve]

We now destroy articles in the background 😄