Red Hat Bugzilla – Bug 1478364
kill passenger threads that have not returned in 60 minutes
Last modified: 2018-04-02 01:19:15 EDT
Description of problem:
Users occasionally hit bugs where the passenger process does not return for hours. While these bugs are serious and need to be addressed, it would be better for the Satellite if we just killed the process after some number of minutes instead of waiting forever.
It would be good if this setting is configurable via custom-hiera.yaml, and if information about the request (url, http action, params, stack trace) is logged when the process is killed.
Version-Release number of selected component (if applicable): 6.2.10
Steps to Reproduce:
1. make an API call that takes an extremely long time to return, or add a sleep statement to induce this behavior
Actual results: The request will eventually return, even after hours. During this time, the passenger processes is not accepting requests.
Expected results: Passenger process gives up after one hour
Additional info: I chose 60 minutes as a high first step since users occasionally run extremely long-running hammer commands. These long-running commands should not take as long, but we can work on lowering the 60 minute threshold as we fix issues.