Safety concern to use thread_queue delivery method in Resque and Sidekiq

[arthurchui]

We use Ruby MRI to run Resque and Sidekiq. Sometimes, our code sends reports to Bugsnag within active db transaction blocks. To avoid long transactions, we prefer to defer the API calls after committing the db transactions.

Since v2.6 (#180), Bugsnag chose synchronous method for Resque and Sidekiq. #180 pointed out that Resque had an issue with ThreadQueue.

After ~6 years, I wonder if it is safe to turn on thread_queue now.

[abigailbramble]

Hi @arthurchui
Unfortunately, this cannot be threaded. The reason we have chosen synchronous is so that the error report finishes sending before the job terminates. If this does not happen, the termination of the job interrupts the notification being sent to Bugsnag.

[arthurchui]

Hi @phillipsam . Thanks for your reply. I wonder the at_exit hook should send the outstanding reports to Bugsnag before the job terminates, right?

[abigailbramble]

Hi @arthurchui it looks like this is something we could explore when priorities allow. Let us know if you have further comments in the meantime.

[johnkiely1]

This has now been released in v6.17.0